root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-09-23 05:26:19 +00:00
Code Issues Actions 11 Packages Projects Releases Wiki Activity
c780e75305
llama.cpp/llama.cpp

17983 lines
716 KiB
C++
Raw Normal View History

check C++ code with -Wmissing-declarations (#3184)
2023-09-15 19:38:27 +00:00
#define LLAMA_API_INTERNAL
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
#include "llama.h"
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
#include "unicode.h"
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
#include "ggml.h"
ggml-cuda : use graph allocator (#2684) use a different function for no_alloc to avoid breaking backwards compat, fixes lora remove 512 n_batch limit fixed 2048 batch size cleanup Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2023-08-22 13:25:19 +00:00
#include "ggml-alloc.h"
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
#include "ggml-backend.h"
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
cuda : rename build flag to LLAMA_CUDA (#6299)
2024-03-26 00:16:01 +00:00
#ifdef GGML_USE_CUDA
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
# include "ggml-cuda.h"
OpenCL Token Generation Acceleration (#1459) * Move back to C++ for OpenCL * Refactor OpenCL code to work more like the CUDA code, add missing functions * Deduplicate dequant kernels * Add OpenCL compile options * Use compile args for preprocessing constants * Restore default platform + device selection by id behavior --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Henri Vasserman <henv@hot.ee>
2023-05-22 21:33:24 +00:00
#elif defined(GGML_USE_CLBLAST)
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
# include "ggml-opencl.h"
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 17:03:59 +00:00
#elif defined(GGML_USE_VULKAN)
# include "ggml-vulkan.h"
ggml : add unified SYCL backend for Intel GPUs (#2690) * first update for migration * update init_cublas * add debug functio, commit all help code * step 1 * step 2 * step3 add fp16, slower 31->28 * add GGML_LIST_DEVICE function * step 5 format device and print * step6, enhance error check, remove CUDA macro, enhance device id to fix none-zero id issue * support main device is non-zero * step7 add debug for code path, rm log * step 8, rename all macro & func from cuda by sycl * fix error of select non-zero device, format device list * ren ggml-sycl.hpp -> ggml-sycl.h * clear CMAKE to rm unused lib and options * correct queue: rm dtct:get_queue * add print tensor function to debug * fix error: wrong result in 658746bb26702e50f2c59c0e4ada8e9da6010481 * summary dpct definition in one header file to replace folder:dpct * refactor device log * mv dpct definition from folder dpct to ggml-sycl.h * update readme, refactor build script * fix build with sycl * set nthread=1 when sycl, increase performance * add run script, comment debug code * add ls-sycl-device tool * add ls-sycl-device, rm unused files * rm rear space * dos2unix * Update README_sycl.md * fix return type * remove sycl version from include path * restore rm code to fix hang issue * add syc and link for sycl readme * rm original sycl code before refactor * fix code err * add know issue for pvc hang issue * enable SYCL_F16 support * align pr4766 * check for sycl blas, better performance * cleanup 1 * remove extra endif * add build&run script, clean CMakefile, update guide by review comments * rename macro to intel hardware * editor config format * format fixes * format fixes * editor format fix * Remove unused headers * skip build sycl tool for other code path * replace tab by space * fix blas matmul function * fix mac build * restore hip dependency * fix conflict * ren as review comments * mv internal function to .cpp file * export funciton print_sycl_devices(), mv class dpct definition to source file * update CI/action for sycl code, fix CI error of repeat/dup * fix action ID format issue * rm unused strategy * enable llama_f16 in ci * fix conflict * fix build break on MacOS, due to CI of MacOS depend on external ggml, instead of internal ggml * fix ci cases for unsupported data type * revert unrelated changed in cuda cmake remove useless nommq fix typo of GGML_USE_CLBLAS_SYCL * revert hip cmake changes * fix indent * add prefix in func name * revert no mmq * rm cpu blas duplicate * fix no_new_line * fix src1->type==F16 bug. * pass batch offset for F16 src1 * fix batch error * fix wrong code * revert sycl checking in test-sampling * pass void as arguments of ggml_backend_sycl_print_sycl_devices * remove extra blank line in test-sampling * revert setting n_threads in sycl * implement std::isinf for icpx with fast math. * Update ci/run.sh Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update examples/sycl/run-llama2.sh Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update examples/sycl/run-llama2.sh Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * add copyright and MIT license declare * update the cmd example --------- Co-authored-by: jianyuzh <jianyu.zhang@intel.com> Co-authored-by: luoyu-intel <yu.luo@intel.com> Co-authored-by: Meng, Hengyu <hengyu.meng@intel.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 15:56:23 +00:00
#elif defined(GGML_USE_SYCL)
# include "ggml-sycl.h"
Nomic Vulkan backend (#4456) Signed-off-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: niansa <anton-sa@web.de> Co-authored-by: Adam Treat <treat.adam@gmail.com> Co-authored-by: Aaron Miller <apage43@ninjawhale.com> Co-authored-by: ToKiNoBug <tokinobug@163.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-01-29 20:50:50 +00:00
#elif defined(GGML_USE_KOMPUTE)
# include "ggml-kompute.h"
ggml : GPU-accelerated token generation (#1412) * CUDA kernel for q4_0 dequant. + mat. vec. mult. * Added q4_1 via template * Added missing __syncthreads(); * --gpu_layers -> --gpu-layers * Shorter dequantize_mul_mat_vec line * q5_0 dequantize_mul_mat kernel * More readable dequantize_mul_mat_vec logic * dequantize_mul_mat_vec kernels for q5_1, q8_0, f16 * llama : offload "output" tensor to GPU too + coding style fixes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-05-13 13:38:36 +00:00
#endif
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : Metal inference (#1642) * mtl : export the LLaMA computation graph * ci : disable temporary * mtl : adapt the MNIST example as starter * mtl : no need for mtl-export tool, add cli arg for main instead * mtl : export just a small part of the graph for now to make it easier * mtl : move MSL code into separate file for easy editing * mtl : initial get_rows_q4_0 kernel * mtl : confirmed get_rows_q4_0 is working correctly * mtl : add rms_norm kernel + confirm working * mtl : add mul kernel + confirm working * mtl : initial mul_mat Q4 kernel (wrong results) * mtl : mul_mat fixes (still wrong) * mtl : another mul_mat Q4 (still does not work) * mtl : working mul_mat q4 * ggml : fix handling of "view" ops in ggml_graph_import() * mtl : add rope kernel * mtl : add reshape and transpose handling * ggml : store offset as opt arg for ggml_view_xd() operators * mtl : add cpy kernel + handle view ops * mtl : confirm f16 x f32 attention mul mat * mtl : add scale kernel * mtl : add diag_mask_inf kernel * mtl : fix soft_max kernel * ggml : update ggml_nbytes() to handle non-contiguous tensors * mtl : verify V tensor contents * mtl : add f32 -> f32 cpy kernel * mtl : add silu kernel * mtl : add non-broadcast mul kernel * mtl : full GPU inference of the computation graph * mtl : optimize rms_norm and soft_max kernels * mtl : add f16 mat x f32 vec multiplication kernel * mtl : fix bug in f16 x f32 mul mat + speed-up computation * mtl : faster mul_mat_q4_0_f32 kernel * mtl : fix kernel signature + roll inner loop * mtl : more threads for rms_norm + better timing * mtl : remove printfs from inner loop * mtl : simplify implementation * mtl : add save/load vocab to ggml file * mtl : plug Metal inference into llama.cpp (very quick-n-dirty) * mtl : make it work with main example Lots of hacks but at least now it generates text * mtl : preparing for merge * mtl : clean-up ggml mtl interface + suport scratch / inplace * mtl : remove temp / debug code * metal : final refactoring and simplification * Revert "ci : disable temporary" This reverts commit 98c267fc77fe811082f672538fc91bcfc9072d63. * metal : add comments * metal : clean-up stuff, fix typos * readme : add Metal instructions * readme : add example for main
2023-06-04 20:34:30 +00:00
#ifdef GGML_USE_METAL
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
# include "ggml-metal.h"
llama : Metal inference (#1642) * mtl : export the LLaMA computation graph * ci : disable temporary * mtl : adapt the MNIST example as starter * mtl : no need for mtl-export tool, add cli arg for main instead * mtl : export just a small part of the graph for now to make it easier * mtl : move MSL code into separate file for easy editing * mtl : initial get_rows_q4_0 kernel * mtl : confirmed get_rows_q4_0 is working correctly * mtl : add rms_norm kernel + confirm working * mtl : add mul kernel + confirm working * mtl : initial mul_mat Q4 kernel (wrong results) * mtl : mul_mat fixes (still wrong) * mtl : another mul_mat Q4 (still does not work) * mtl : working mul_mat q4 * ggml : fix handling of "view" ops in ggml_graph_import() * mtl : add rope kernel * mtl : add reshape and transpose handling * ggml : store offset as opt arg for ggml_view_xd() operators * mtl : add cpy kernel + handle view ops * mtl : confirm f16 x f32 attention mul mat * mtl : add scale kernel * mtl : add diag_mask_inf kernel * mtl : fix soft_max kernel * ggml : update ggml_nbytes() to handle non-contiguous tensors * mtl : verify V tensor contents * mtl : add f32 -> f32 cpy kernel * mtl : add silu kernel * mtl : add non-broadcast mul kernel * mtl : full GPU inference of the computation graph * mtl : optimize rms_norm and soft_max kernels * mtl : add f16 mat x f32 vec multiplication kernel * mtl : fix bug in f16 x f32 mul mat + speed-up computation * mtl : faster mul_mat_q4_0_f32 kernel * mtl : fix kernel signature + roll inner loop * mtl : more threads for rms_norm + better timing * mtl : remove printfs from inner loop * mtl : simplify implementation * mtl : add save/load vocab to ggml file * mtl : plug Metal inference into llama.cpp (very quick-n-dirty) * mtl : make it work with main example Lots of hacks but at least now it generates text * mtl : preparing for merge * mtl : clean-up ggml mtl interface + suport scratch / inplace * mtl : remove temp / debug code * metal : final refactoring and simplification * Revert "ci : disable temporary" This reverts commit 98c267fc77fe811082f672538fc91bcfc9072d63. * metal : add comments * metal : clean-up stuff, fix typos * readme : add Metal instructions * readme : add example for main
2023-06-04 20:34:30 +00:00
#endif
mpi : add support for distributed inference via MPI (#2099) * MPI support, first cut * fix warnings, update README * fixes * wrap includes * PR comments * Update CMakeLists.txt * Add GH workflow, fix test * Add info to README * mpi : trying to move more MPI stuff into ggml-mpi (WIP) (#2099) * mpi : add names for layer inputs + prep ggml_mpi_graph_compute() * mpi : move all MPI logic into ggml-mpi Not tested yet * mpi : various fixes - communication now works but results are wrong * mpi : fix output tensor after MPI compute (still not working) * mpi : fix inference * mpi : minor * Add OpenMPI to GH action * [mpi] continue-on-error: true * mpi : fix after master merge * [mpi] Link MPI C++ libraries to fix OpenMPI * tests : fix new llama_backend API * [mpi] use MPI_INT32_T * mpi : factor out recv / send in functions and reuse * mpi : extend API to allow usage with outer backends (e.g. Metal) --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-10 15:49:56 +00:00
#ifdef GGML_USE_MPI
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
# include "ggml-mpi.h"
mpi : add support for distributed inference via MPI (#2099) * MPI support, first cut * fix warnings, update README * fixes * wrap includes * PR comments * Update CMakeLists.txt * Add GH workflow, fix test * Add info to README * mpi : trying to move more MPI stuff into ggml-mpi (WIP) (#2099) * mpi : add names for layer inputs + prep ggml_mpi_graph_compute() * mpi : move all MPI logic into ggml-mpi Not tested yet * mpi : various fixes - communication now works but results are wrong * mpi : fix output tensor after MPI compute (still not working) * mpi : fix inference * mpi : minor * Add OpenMPI to GH action * [mpi] continue-on-error: true * mpi : fix after master merge * [mpi] Link MPI C++ libraries to fix OpenMPI * tests : fix new llama_backend API * [mpi] use MPI_INT32_T * mpi : factor out recv / send in functions and reuse * mpi : extend API to allow usage with outer backends (e.g. Metal) --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-10 15:49:56 +00:00
#endif
ggml : quantization refactoring (#3833) * ggml : factor all quantization code in ggml-quants ggml-ci * ggml-quants : fix Zig and Swift builds + quantize tool ggml-ci * quantize : --pure option for disabling k-quant mixtures --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com>
2023-10-29 16:32:28 +00:00
#ifndef QK_K
# ifdef GGML_QKK_64
# define QK_K 64
# else
# define QK_K 256
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
# endif
llama : prevent usage of k-quants when tensor size is not a multiple of 256 (#1921) * Fix examples/metal * k-quants: prevent usage when tensor size is not divisible by 256 --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2023-06-18 08:13:43 +00:00
#endif
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#ifdef __has_include
#if __has_include(<unistd.h>)
#include <unistd.h>
#if defined(_POSIX_MAPPED_FILES)
#include <sys/mman.h>
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
#include <fcntl.h>
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#endif
#if defined(_POSIX_MEMLOCK_RANGE)
#include <sys/resource.h>
#endif
#endif
llama : prevent usage of k-quants when tensor size is not a multiple of 256 (#1921) * Fix examples/metal * k-quants: prevent usage when tensor size is not divisible by 256 --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2023-06-18 08:13:43 +00:00
#endif
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#if defined(_WIN32)
#define WIN32_LEAN_AND_MEAN
#ifndef NOMINMAX
#define NOMINMAX
#endif
#include <windows.h>
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
#ifndef PATH_MAX
#define PATH_MAX MAX_PATH
#endif
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#include <io.h>
k-quants : support for super-block size of 64 (#2001) * k_quants: WIP super-blocks with 64 weights * k_quants: WIP super-blocks with 64 weights Q6_K scalar and AVX2 works * k_quants: WIP super-blocks with 64 weights Q4_K scalar and AVX2 works * k_quants: WIP super-blocks with 64 weights Q2_K scalar and AVX2 works. Q2_K is way too slow (it is actually slower than the scalar implementation) * k_quants: WIP super-blocks with 64 weights Q3_K scalar and AVX2 works. * k_quants: WIP super-blocks with 64 weights Q5_K scalar and AVX2 works, and with that all k_quants are done on AVX2 and scalar * k_quants: WIP super-blocks with 64 weights Q6_K working on CUDA. Cannot make it run quite as gast as with super-blocks with 256 weigths: 8% slower on 4080, 20% slower on the 1660 (but there we fit 1 less layer on the GPU because pf the larger model size), so some fraction of these 20% is due to that, * k_quants: WIP super-blocks with 64 weights Q4_K working on CUDA. ~10% slower on GTX-1660, 16% slower on 4080. * k_quants: WIP super-blocks with 64 weights Q2_K working on CUDA. ~3% slower on GTX-1660, 10% slower on 4080. * k_quants: WIP super-blocks with 64 weights Q3_K working on CUDA. * k_quants: WIP super-blocks with 64 weights Q5_K working on CUDA, and with this CUDA is done. * k_quants: WIP super-blocks with 64 weights Q6_K working on ARM_NEON * k_quants: WIP super-blocks with 64 weights Q4_K working on ARM_NEON, but quite a bit slower than 256 weights * k_quants: WIP super-blocks with 64 weights Q2_K working on ARM_NEON, but quite a bit slower than 256 weights * k_quants: WIP super-blocks with 64 weights Q3_K working on ARM_NEON, but quite a bit slower than 256 weights. * k_quants: WIP super-blocks with 64 weights Q5_K working on ARM_NEON, but quite a bit slower than 256 weights. With that, we have full support for ARM_NEON, although performance is not quite there. * k_quants: WIP super-blocks with 64 weights Slightly more efficient Q3_K and Q5_K * k_quants: WIP super-blocks with 64 weights Another small improvement for Q3_K and Q5_K on ARM_NEON * k_quants: WIP super-blocks with 64 weights Yet another speedup for Q5_K on ARM_NEON. We are now within 10% of the QK_K = 256 version. * k_quants: WIP super-blocks with 64 weights * We are able to pass preprocessor macros to the Metal compiler * Q6_K works and is actually slightly more efficient than the QK_K = 256 version (25.2 ms vs 25.8 ms) * k_quants: WIP super-blocks with 64 weights Q4_K works on Metal and is actually slightly faster than QK_K = 256 (21.95 ms vs 24.0 ms). * k_quants: WIP super-blocks with 64 weights Q2_K works on Metal and is very slightly faster than QK_K = 256 (23.8 ms vs 24.2 ms). * k_quants: WIP super-blocks with 64 weights Q3_K works on Metal and is slightly faster than QK_K = 256 (26.6 ms vs 28.3 ms). * k_quants: WIP super-blocks with 64 weights Q5_K works on Metal and is slightly faster than QK_K = 256 (23.7 ms vs 26.3 ms). * k_quants: call them _K, not _k, also on Metal * k_quants: correctly define QK_K in llama.cpp * Fixed bug in q4_K quantization added with the 64-block addition * Simplify via lambda * k_quants: swicth Q3_K to 4-bit scales when QK_K = 64 Otherwise there isn't much benefit from this quantization type. There is some very slight loss in accuracy, but we reduce size by ~7%. E.g., for OpenLLaMA-3B, Q3_K_S perplexity is 8.6131 with 8-bit scales and 8.6352 with 4-bit, while file size decreases from 1.53G to 1.44G. * k_quants: switch Q4_K to 4-bit scales when QK_K = 64 Here the loss in accuracy is greater than for Q3_K, but the Q4_K points still move further to the left on the perplexity vs size curve. * k_quants: forgot to add the Metal changes in last commit * k_quants: change Q5_K to be type 0 when QK_K = 64 Still needs AVX2 implementation * k_quants: AVX2 implementation for new 64-weight Q5_K * k_quants: 10% faster ARM_NEON Q5_K dot product * k_quants: fixed issue caused by merging with master --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2023-06-26 16:43:07 +00:00
#endif
llama : Metal inference (#1642) * mtl : export the LLaMA computation graph * ci : disable temporary * mtl : adapt the MNIST example as starter * mtl : no need for mtl-export tool, add cli arg for main instead * mtl : export just a small part of the graph for now to make it easier * mtl : move MSL code into separate file for easy editing * mtl : initial get_rows_q4_0 kernel * mtl : confirmed get_rows_q4_0 is working correctly * mtl : add rms_norm kernel + confirm working * mtl : add mul kernel + confirm working * mtl : initial mul_mat Q4 kernel (wrong results) * mtl : mul_mat fixes (still wrong) * mtl : another mul_mat Q4 (still does not work) * mtl : working mul_mat q4 * ggml : fix handling of "view" ops in ggml_graph_import() * mtl : add rope kernel * mtl : add reshape and transpose handling * ggml : store offset as opt arg for ggml_view_xd() operators * mtl : add cpy kernel + handle view ops * mtl : confirm f16 x f32 attention mul mat * mtl : add scale kernel * mtl : add diag_mask_inf kernel * mtl : fix soft_max kernel * ggml : update ggml_nbytes() to handle non-contiguous tensors * mtl : verify V tensor contents * mtl : add f32 -> f32 cpy kernel * mtl : add silu kernel * mtl : add non-broadcast mul kernel * mtl : full GPU inference of the computation graph * mtl : optimize rms_norm and soft_max kernels * mtl : add f16 mat x f32 vec multiplication kernel * mtl : fix bug in f16 x f32 mul mat + speed-up computation * mtl : faster mul_mat_q4_0_f32 kernel * mtl : fix kernel signature + roll inner loop * mtl : more threads for rms_norm + better timing * mtl : remove printfs from inner loop * mtl : simplify implementation * mtl : add save/load vocab to ggml file * mtl : plug Metal inference into llama.cpp (very quick-n-dirty) * mtl : make it work with main example Lots of hacks but at least now it generates text * mtl : preparing for merge * mtl : clean-up ggml mtl interface + suport scratch / inplace * mtl : remove temp / debug code * metal : final refactoring and simplification * Revert "ci : disable temporary" This reverts commit 98c267fc77fe811082f672538fc91bcfc9072d63. * metal : add comments * metal : clean-up stuff, fix typos * readme : add Metal instructions * readme : add example for main
2023-06-04 20:34:30 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#include <algorithm>
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
#include <array>
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#include <cassert>
wpm : portable unicode tolower (#6305) Also use C locale for ispunct/isspace, and split unicode-data.cpp from unicode.cpp.
2024-03-26 21:46:21 +00:00
#include <cctype>
Apply min_p to unsorted tokens (#5115)
2024-01-28 08:59:49 +00:00
#include <cfloat>
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
#include <cinttypes>
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#include <climits>
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
#include <cmath>
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#include <cstdarg>
llama : move #includes out of _GNU_SOURCE conditional (#2817)
2023-08-26 18:17:51 +00:00
#include <cstddef>
#include <cstdint>
#include <cstdio>
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#include <cstring>
#include <ctime>
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
#include <forward_list>
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
#include <fstream>
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
#include <functional>
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
#include <future>
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#include <initializer_list>
llama : improve BERT tokenization (#5740) * implement nfd for stripping accents in wpm tokenizer * sort nfd map; reuse iterator * use builtin tolower * add locale include * Simplify to_lower cases Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-02-28 08:51:11 +00:00
#include <locale>
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
#include <map>
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
#include <memory>
llama : multi-threaded quantization (#1075) * Multi-threading quantization. Not much gain for simple quantizations, bit it will be important for quantizations that require more CPU cycles. * Multi-threading for quantize-stats It now does the job in ~14 seconds on my Mac for Q4_0, Q4_1 and Q4_2. Single-threaded it was taking more than 2 minutes after adding the more elaborate version of Q4_2. * Reviewer comments * Avoiding compiler confusion After changing chunk_size to const int as suggested by @ggerganov, clang and GCC starting to warn me that I don't need to capture it in the lambda. So, I removed it from the capture list. But that makes the MSVC build fail. So, making it a constexpr to make every compiler happy. * Still fighting with lambda captures in MSVC --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-04-20 17:42:27 +00:00
#include <mutex>
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
#include <numeric>
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#include <queue>
#include <random>
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
#include <regex>
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
#include <set>
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#include <sstream>
#include <thread>
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
#include <type_traits>
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#include <unordered_map>
Make loading weights 10-100x faster This is a breaking change that's going to give you three benefits: 1. Your inference commands should load 100x faster 2. You may be able to safely load models 2x larger 3. You can run many concurrent inference processes This was accomplished by changing the file format so we can mmap() weights directly into memory without having to read() or copy them thereby ensuring the kernel can make its file cache pages directly accessible to our inference processes; and secondly, that the file cache pages are much less likely to get evicted (which would force loads to hit disk) because they're no longer competing with memory pages that were needlessly created by gigabytes of standard i/o. The new file format supports single-file models like LLaMA 7b, and it also supports multi-file models like LLaMA 13B. Our Python tool now merges the foo.1, foo.2, etc. files back into a single file so that the C++ code which maps it doesn't need to reshape data every time. That's made llama.cpp so much simpler. Much of its load code has now been deleted. Furthermore, this change ensures that tensors are aligned properly on a 32-byte boundary. That opens the door to seeing if we can get additional performance gains on some microprocessors, by using ops that require memory alignment. Lastly note that both POSIX and the Windows platform are supported Fixes #91
2023-03-29 20:51:37 +00:00
build : fix and ignore MSVC warnings (#1889)
2023-06-16 18:23:53 +00:00
#if defined(_MSC_VER)
#pragma warning(disable: 4244 4267) // possible loss of data
#endif
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#ifdef __GNUC__
#ifdef __MINGW32__
#define LLAMA_ATTRIBUTE_FORMAT(...) __attribute__((format(gnu_printf, __VA_ARGS__)))
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
#else
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#define LLAMA_ATTRIBUTE_FORMAT(...) __attribute__((format(printf, __VA_ARGS__)))
#endif
#else
#define LLAMA_ATTRIBUTE_FORMAT(...)
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
#endif
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
#define LLAMA_MAX_NODES 8192
llama : add qwen2moe (#6074) * support qwen2moe * fix-review * metal : support unary ops for nelements % 4 != 0 * metal : require contiguousness for float4 unary kernels * metal : require contiguousness for float4 unary kernels (cont) * fix-review * names : for brevity "SHARED_EXP" -> "SHEXP" * llama : reuse build_moe_ffn() * llama : add model type name --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-16 15:40:48 +00:00
#define LLAMA_MAX_EXPERTS 60
sync : ggml (backend v2) (#3912) * sync : ggml (backend v2) (wip) * sync : migrate examples and llama.cpp to dynamic graphs (wip) * sync : update tests + fix max op params to 64 ggml-ci * sync : ggml-cuda ggml-ci * llama : fix save/load state context size ggml-ci * sync : try to fix build on tvOS * sync : pass custom graph sizes in training examples * sync : update graph copies to new ggml API * sync : update sync-ggml.sh with new files * scripts : fix header in sync script * train : fix context size calculations * llama : increase inference graph size up to 4096 nodes * train : allocate grads for backward graphs * train : allocate grads for gb_tmp
2023-11-13 12:16:23 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
//
// logging
//
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
LLAMA_ATTRIBUTE_FORMAT(2, 3)
metal : reusing llama.cpp logging (#3152) * metal : reusing llama.cpp logging * cmake : build fix * metal : logging callback * metal : logging va_args memory fix * metal : minor cleanup * metal : setting function like logging macro to capital letters * llama.cpp : trailing whitespace fix * ggml : log level enum used by llama * Makefile : cleanup ggml-metal recipe * ggml : ggml_log_callback typedef * ggml : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-27 15:48:33 +00:00
static void llama_log_internal (ggml_log_level level, const char* format, ...);
static void llama_log_callback_default(ggml_log_level level, const char * text, void * user_data);
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
metal : reusing llama.cpp logging (#3152) * metal : reusing llama.cpp logging * cmake : build fix * metal : logging callback * metal : logging va_args memory fix * metal : minor cleanup * metal : setting function like logging macro to capital letters * llama.cpp : trailing whitespace fix * ggml : log level enum used by llama * Makefile : cleanup ggml-metal recipe * ggml : ggml_log_callback typedef * ggml : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-27 15:48:33 +00:00
#define LLAMA_LOG_INFO(...) llama_log_internal(GGML_LOG_LEVEL_INFO , __VA_ARGS__)
#define LLAMA_LOG_WARN(...) llama_log_internal(GGML_LOG_LEVEL_WARN , __VA_ARGS__)
#define LLAMA_LOG_ERROR(...) llama_log_internal(GGML_LOG_LEVEL_ERROR, __VA_ARGS__)
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
//
// helpers
//
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
static size_t utf8_len(char src) {
const size_t lookup[] = { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 3, 4 };
uint8_t highbits = static_cast<uint8_t>(src) >> 4;
return lookup[highbits];
}
check C++ code with -Wmissing-declarations (#3184)
2023-09-15 19:38:27 +00:00
static void replace_all(std::string & s, const std::string & search, const std::string & replace) {
llama : speedup tokenization (#2831) * Speedup tokenization On current master it takes ~3.2 seconds to tokenize Wikitext. With this change it becomes ~525 ms. * Fixit: it was missing the piece after the last found occurence --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2023-08-27 13:50:33 +00:00
std::string result;
for (size_t pos = 0; ; pos += search.length()) {
auto new_pos = s.find(search, pos);
if (new_pos == std::string::npos) {
result += s.substr(pos, s.size() - pos);
break;
}
result += s.substr(pos, new_pos - pos) + replace;
pos = new_pos;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
llama : speedup tokenization (#2831) * Speedup tokenization On current master it takes ~3.2 seconds to tokenize Wikitext. With this change it becomes ~525 ms. * Fixit: it was missing the piece after the last found occurence --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2023-08-27 13:50:33 +00:00
s = std::move(result);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
llama : correct hparams comparison (#3446) * fixed floating point comparison issues * updated implementation for hparam comparison to handle inf and NaN * fixed code review comments * minor simplification * rename is_float_eq -> is_float_close --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com>
2023-10-06 10:47:59 +00:00
static bool is_float_close(float a, float b, float abs_tol) {
// Check for non-negative tolerance
if (abs_tol < 0.0) {
throw std::invalid_argument("Tolerance must be non-negative");
}
// Exact equality check
if (a == b) {
return true;
}
// Check for infinities
if (std::isinf(a) || std::isinf(b)) {
return false;
}
// Regular comparison using the provided absolute tolerance
return std::fabs(b - a) <= abs_tol;
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
static void zeros(std::ofstream & file, size_t n) {
char zero = 0;
for (size_t i = 0; i < n; ++i) {
file.write(&zero, 1);
}
}
Multi GPU support, CUDA refactor, CUDA scratch buffer (#1703) * CUDA multi GPU + scratch ggml_cuda_compute_forward Tensor parallelism ggml_cuda_add ggml_cuda_rms_norm ggml_cuda_silu CUDA scratch buffer --main-gpu CLI option
2023-06-06 19:33:23 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
LLAMA_ATTRIBUTE_FORMAT(1, 2)
static std::string format(const char * fmt, ...) {
va_list ap;
va_list ap2;
va_start(ap, fmt);
va_copy(ap2, ap);
int size = vsnprintf(NULL, 0, fmt, ap);
GGML_ASSERT(size >= 0 && size < INT_MAX); // NOLINT
std::vector<char> buf(size + 1);
int size2 = vsnprintf(buf.data(), size + 1, fmt, ap2);
GGML_ASSERT(size2 == size);
va_end(ap2);
va_end(ap);
return std::string(buf.data(), size);
Multi GPU support, CUDA refactor, CUDA scratch buffer (#1703) * CUDA multi GPU + scratch ggml_cuda_compute_forward Tensor parallelism ggml_cuda_add ggml_cuda_rms_norm ggml_cuda_silu CUDA scratch buffer --main-gpu CLI option
2023-06-06 19:33:23 +00:00
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
//
// gguf constants (sync with gguf.py)
//
enum llm_arch {
LLM_ARCH_LLAMA,
LLM_ARCH_FALCON,
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
LLM_ARCH_BAICHUAN,
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
LLM_ARCH_GROK,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
LLM_ARCH_GPT2,
LLM_ARCH_GPTJ,
LLM_ARCH_GPTNEOX,
LLM_ARCH_MPT,
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
LLM_ARCH_STARCODER,
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
LLM_ARCH_PERSIMMON,
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
LLM_ARCH_REFACT,
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
LLM_ARCH_BERT,
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
LLM_ARCH_NOMIC_BERT,
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
LLM_ARCH_BLOOM,
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
LLM_ARCH_STABLELM,
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
LLM_ARCH_QWEN,
llama : support upcoming Qwen2 (#5037)
2024-01-19 11:53:13 +00:00
LLM_ARCH_QWEN2,
llama : add qwen2moe (#6074) * support qwen2moe * fix-review * metal : support unary ops for nelements % 4 != 0 * metal : require contiguousness for float4 unary kernels * metal : require contiguousness for float4 unary kernels (cont) * fix-review * names : for brevity "SHARED_EXP" -> "SHEXP" * llama : reuse build_moe_ffn() * llama : add model type name --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-16 15:40:48 +00:00
LLM_ARCH_QWEN2MOE,
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
LLM_ARCH_PHI2,
llama : add phi3 support (#6852) * add explicit phi3 support * add explicit phi3 support * remove unused code * convert : add BOS token * llama : match EOT token <|end|> * llama : minor / style * llama : tabs -> spaces * convert : fix lint checks --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-24 07:00:37 +00:00
LLM_ARCH_PHI3,
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
LLM_ARCH_PLAMO,
llama : add CodeShell support (#5016) * llama: add codeshell support * llama.cpp: fix codeshell with NeoX rope Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-19 09:07:27 +00:00
LLM_ARCH_CODESHELL,
llama : add support for Orion-14B (#5118) * add support for Orion-14B(https://huggingface.co/OrionStarAI/Orion-14B-Chat) * flake8 support * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * Update llama.cpp * Update llama.cpp --------- Co-authored-by: lixiaopu <lixiaopu@cmcm.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-01-28 08:00:30 +00:00
LLM_ARCH_ORION,
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
LLM_ARCH_INTERNLM2,
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
LLM_ARCH_MINICPM,
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
LLM_ARCH_GEMMA,
llama : add StarCoder2 support (#5795) * Add support for starcoder2 * handle rope type * skip rope freq and rotary embeddings from being serialized * resolve comments * Update llama.cpp * remove redundant changes * handle `rope-theta` * llama : change starcoder2 rope type * address comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-01 19:30:46 +00:00
LLM_ARCH_STARCODER2,
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
LLM_ARCH_MAMBA,
[Model] Add support for xverse (#6301) * Support xverse model convert to gguf format. * 1. Convert xverse models to gguf; 2. Add LLM_ARCH_XVERSE inference in llama.cpp; 3. Add xverse item in Supported models in README.md; * * gguf-py: remove redundant logs * llama: remove the init_mapping_prefetch custom parameter * llama.cpp: Include the changes from #6122 to exclude the unused outputs of the last layers. * - Fix format issues - Remove duplicate set kqv_out to llm_build_kv * Update llama.cpp --------- Co-authored-by: willhe <willhe@xverse.cn> Co-authored-by: willhe <hexin@xverse.cn>
2024-03-29 13:37:03 +00:00
LLM_ARCH_XVERSE,
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
LLM_ARCH_COMMAND_R,
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
LLM_ARCH_DBRX,
Implement the OLMo architecture (#6741) * implement olmo architecture * remove unused variable * remove unused moe branch * remove check for weight * remove superfluous moe, bias and rope tensors * clarified comment * fix clamp_kqv setting * remove obsolete parameter name filter
2024-04-19 09:35:54 +00:00
LLM_ARCH_OLMO,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
LLM_ARCH_UNKNOWN,
};
llama : fix segfault from unknown model arch name (#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion https://github.com/ggerganov/llama.cpp/pull/5820#issuecomment-1973735284 * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-03-02 13:42:56 +00:00
static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
{ LLM_ARCH_LLAMA, "llama" },
{ LLM_ARCH_FALCON, "falcon" },
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
{ LLM_ARCH_GROK, "grok" },
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
{ LLM_ARCH_GPT2, "gpt2" },
{ LLM_ARCH_GPTJ, "gptj" },
{ LLM_ARCH_GPTNEOX, "gptneox" },
{ LLM_ARCH_MPT, "mpt" },
{ LLM_ARCH_BAICHUAN, "baichuan" },
{ LLM_ARCH_STARCODER, "starcoder" },
{ LLM_ARCH_PERSIMMON, "persimmon" },
{ LLM_ARCH_REFACT, "refact" },
{ LLM_ARCH_BERT, "bert" },
{ LLM_ARCH_NOMIC_BERT, "nomic-bert" },
{ LLM_ARCH_BLOOM, "bloom" },
{ LLM_ARCH_STABLELM, "stablelm" },
{ LLM_ARCH_QWEN, "qwen" },
{ LLM_ARCH_QWEN2, "qwen2" },
llama : add qwen2moe (#6074) * support qwen2moe * fix-review * metal : support unary ops for nelements % 4 != 0 * metal : require contiguousness for float4 unary kernels * metal : require contiguousness for float4 unary kernels (cont) * fix-review * names : for brevity "SHARED_EXP" -> "SHEXP" * llama : reuse build_moe_ffn() * llama : add model type name --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-16 15:40:48 +00:00
{ LLM_ARCH_QWEN2MOE, "qwen2moe" },
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
{ LLM_ARCH_PHI2, "phi2" },
llama : add phi3 support (#6852) * add explicit phi3 support * add explicit phi3 support * remove unused code * convert : add BOS token * llama : match EOT token <|end|> * llama : minor / style * llama : tabs -> spaces * convert : fix lint checks --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-24 07:00:37 +00:00
{ LLM_ARCH_PHI3, "phi3" },
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
{ LLM_ARCH_PLAMO, "plamo" },
{ LLM_ARCH_CODESHELL, "codeshell" },
{ LLM_ARCH_ORION, "orion" },
{ LLM_ARCH_INTERNLM2, "internlm2" },
{ LLM_ARCH_MINICPM, "minicpm" },
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
{ LLM_ARCH_GEMMA, "gemma" },
llama : add StarCoder2 support (#5795) * Add support for starcoder2 * handle rope type * skip rope freq and rotary embeddings from being serialized * resolve comments * Update llama.cpp * remove redundant changes * handle `rope-theta` * llama : change starcoder2 rope type * address comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-01 19:30:46 +00:00
{ LLM_ARCH_STARCODER2, "starcoder2" },
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
{ LLM_ARCH_MAMBA, "mamba" },
[Model] Add support for xverse (#6301) * Support xverse model convert to gguf format. * 1. Convert xverse models to gguf; 2. Add LLM_ARCH_XVERSE inference in llama.cpp; 3. Add xverse item in Supported models in README.md; * * gguf-py: remove redundant logs * llama: remove the init_mapping_prefetch custom parameter * llama.cpp: Include the changes from #6122 to exclude the unused outputs of the last layers. * - Fix format issues - Remove duplicate set kqv_out to llm_build_kv * Update llama.cpp --------- Co-authored-by: willhe <willhe@xverse.cn> Co-authored-by: willhe <hexin@xverse.cn>
2024-03-29 13:37:03 +00:00
{ LLM_ARCH_XVERSE, "xverse" },
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
{ LLM_ARCH_COMMAND_R, "command-r" },
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
{ LLM_ARCH_DBRX, "dbrx" },
Implement the OLMo architecture (#6741) * implement olmo architecture * remove unused variable * remove unused moe branch * remove check for weight * remove superfluous moe, bias and rope tensors * clarified comment * fix clamp_kqv setting * remove obsolete parameter name filter
2024-04-19 09:35:54 +00:00
{ LLM_ARCH_OLMO, "olmo" },
llama : fix segfault from unknown model arch name (#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion https://github.com/ggerganov/llama.cpp/pull/5820#issuecomment-1973735284 * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-03-02 13:42:56 +00:00
{ LLM_ARCH_UNKNOWN, "(unknown)" },
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
};
enum llm_kv {
LLM_KV_GENERAL_ARCHITECTURE,
LLM_KV_GENERAL_QUANTIZATION_VERSION,
LLM_KV_GENERAL_ALIGNMENT,
LLM_KV_GENERAL_NAME,
LLM_KV_GENERAL_AUTHOR,
gguf.py : add licence and version to gguf writer (#6504)
2024-04-05 18:41:38 +00:00
LLM_KV_GENERAL_VERSION,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
LLM_KV_GENERAL_URL,
LLM_KV_GENERAL_DESCRIPTION,
LLM_KV_GENERAL_LICENSE,
LLM_KV_GENERAL_SOURCE_URL,
LLM_KV_GENERAL_SOURCE_HF_REPO,
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
LLM_KV_VOCAB_SIZE,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
LLM_KV_CONTEXT_LENGTH,
LLM_KV_EMBEDDING_LENGTH,
LLM_KV_BLOCK_COUNT,
LLM_KV_FEED_FORWARD_LENGTH,
LLM_KV_USE_PARALLEL_RESIDUAL,
LLM_KV_TENSOR_DATA_LAYOUT,
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
LLM_KV_EXPERT_COUNT,
LLM_KV_EXPERT_USED_COUNT,
Use correct type of pooling for embedding models (#5500) Use correct type of pooling for embedding models
2024-02-15 17:21:49 +00:00
LLM_KV_POOLING_TYPE,
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
LLM_KV_LOGIT_SCALE,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
LLM_KV_ATTENTION_HEAD_COUNT,
LLM_KV_ATTENTION_HEAD_COUNT_KV,
LLM_KV_ATTENTION_MAX_ALIBI_BIAS,
LLM_KV_ATTENTION_CLAMP_KQV,
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
LLM_KV_ATTENTION_KEY_LENGTH,
LLM_KV_ATTENTION_VALUE_LENGTH,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
LLM_KV_ATTENTION_LAYERNORM_EPS,
LLM_KV_ATTENTION_LAYERNORM_RMS_EPS,
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
LLM_KV_ATTENTION_CAUSAL,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
LLM_KV_ROPE_DIMENSION_COUNT,
gguf : add rope_freq_base parameter for CodeLlama (#2769)
2023-08-24 18:04:05 +00:00
LLM_KV_ROPE_FREQ_BASE,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
LLM_KV_ROPE_SCALE_LINEAR,
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
LLM_KV_ROPE_SCALING_TYPE,
LLM_KV_ROPE_SCALING_FACTOR,
LLM_KV_ROPE_SCALING_ORIG_CTX_LEN,
LLM_KV_ROPE_SCALING_FINETUNED,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
LLM_KV_SPLIT_NO,
LLM_KV_SPLIT_COUNT,
LLM_KV_SPLIT_TENSORS_COUNT,
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
LLM_KV_SSM_INNER_SIZE,
LLM_KV_SSM_CONV_KERNEL,
LLM_KV_SSM_STATE_SIZE,
LLM_KV_SSM_TIME_STEP_RANK,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
LLM_KV_TOKENIZER_MODEL,
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
LLM_KV_TOKENIZER_PRE,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
LLM_KV_TOKENIZER_LIST,
LLM_KV_TOKENIZER_TOKEN_TYPE,
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
LLM_KV_TOKENIZER_TOKEN_TYPE_COUNT,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
LLM_KV_TOKENIZER_SCORES,
LLM_KV_TOKENIZER_MERGES,
LLM_KV_TOKENIZER_BOS_ID,
LLM_KV_TOKENIZER_EOS_ID,
LLM_KV_TOKENIZER_UNK_ID,
LLM_KV_TOKENIZER_SEP_ID,
LLM_KV_TOKENIZER_PAD_ID,
BERT tokenizer fixes (#6498) Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
2024-04-09 17:44:08 +00:00
LLM_KV_TOKENIZER_CLS_ID,
LLM_KV_TOKENIZER_MASK_ID,
Respect tokenizer.ggml.add_bos_token value when tokenizing (#4040) * gguf-py: gguf-dump: Respect --no-tensor flag in JSON mode. * Respect add_bos_token GGUF metadata value * gguf-py: Try to fix SpecialVocab giving up too easily for the Nth time
2023-11-17 02:14:37 +00:00
LLM_KV_TOKENIZER_ADD_BOS,
LLM_KV_TOKENIZER_ADD_EOS,
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
LLM_KV_TOKENIZER_ADD_PREFIX,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
LLM_KV_TOKENIZER_HF_JSON,
LLM_KV_TOKENIZER_RWKV,
gguf : add special tokens metadata for FIM/Infill (#6689) This commit adds special token metadata for Fill-In-the-Middle (FIM)/Infill to the GGUF model. The motivation for this is that currently there is support for CodeLlama but other models exist now like CodeGemma, but the different models use different token ids for the special tokens and this commit allows for supporting multiple models. Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
2024-04-16 06:13:13 +00:00
LLM_KV_TOKENIZER_PREFIX_ID,
LLM_KV_TOKENIZER_SUFFIX_ID,
LLM_KV_TOKENIZER_MIDDLE_ID,
LLM_KV_TOKENIZER_EOT_ID,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
};
llama : fix segfault from unknown model arch name (#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion https://github.com/ggerganov/llama.cpp/pull/5820#issuecomment-1973735284 * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-03-02 13:42:56 +00:00
static const std::map<llm_kv, const char *> LLM_KV_NAMES = {
gguf : fix a few general keys (#3341)
2023-09-27 16:18:07 +00:00
{ LLM_KV_GENERAL_ARCHITECTURE, "general.architecture" },
{ LLM_KV_GENERAL_QUANTIZATION_VERSION, "general.quantization_version" },
{ LLM_KV_GENERAL_ALIGNMENT, "general.alignment" },
{ LLM_KV_GENERAL_NAME, "general.name" },
{ LLM_KV_GENERAL_AUTHOR, "general.author" },
gguf.py : add licence and version to gguf writer (#6504)
2024-04-05 18:41:38 +00:00
{ LLM_KV_GENERAL_VERSION, "general.version" },
gguf : fix a few general keys (#3341)
2023-09-27 16:18:07 +00:00
{ LLM_KV_GENERAL_URL, "general.url" },
{ LLM_KV_GENERAL_DESCRIPTION, "general.description" },
{ LLM_KV_GENERAL_LICENSE, "general.license" },
{ LLM_KV_GENERAL_SOURCE_URL, "general.source.url" },
{ LLM_KV_GENERAL_SOURCE_HF_REPO, "general.source.huggingface.repository" },
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
{ LLM_KV_VOCAB_SIZE, "%s.vocab_size" },
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
{ LLM_KV_CONTEXT_LENGTH, "%s.context_length" },
{ LLM_KV_EMBEDDING_LENGTH, "%s.embedding_length" },
{ LLM_KV_BLOCK_COUNT, "%s.block_count" },
{ LLM_KV_FEED_FORWARD_LENGTH, "%s.feed_forward_length" },
{ LLM_KV_USE_PARALLEL_RESIDUAL, "%s.use_parallel_residual" },
{ LLM_KV_TENSOR_DATA_LAYOUT, "%s.tensor_data_layout" },
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
{ LLM_KV_EXPERT_COUNT, "%s.expert_count" },
{ LLM_KV_EXPERT_USED_COUNT, "%s.expert_used_count" },
Use correct type of pooling for embedding models (#5500) Use correct type of pooling for embedding models
2024-02-15 17:21:49 +00:00
{ LLM_KV_POOLING_TYPE , "%s.pooling_type" },
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
{ LLM_KV_LOGIT_SCALE, "%s.logit_scale" },
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
{ LLM_KV_ATTENTION_HEAD_COUNT, "%s.attention.head_count" },
{ LLM_KV_ATTENTION_HEAD_COUNT_KV, "%s.attention.head_count_kv" },
{ LLM_KV_ATTENTION_MAX_ALIBI_BIAS, "%s.attention.max_alibi_bias" },
{ LLM_KV_ATTENTION_CLAMP_KQV, "%s.attention.clamp_kqv" },
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
{ LLM_KV_ATTENTION_KEY_LENGTH, "%s.attention.key_length" },
{ LLM_KV_ATTENTION_VALUE_LENGTH, "%s.attention.value_length" },
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
{ LLM_KV_ATTENTION_LAYERNORM_EPS, "%s.attention.layer_norm_epsilon" },
{ LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, "%s.attention.layer_norm_rms_epsilon" },
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
{ LLM_KV_ATTENTION_CAUSAL, "%s.attention.causal" },
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
{ LLM_KV_ROPE_DIMENSION_COUNT, "%s.rope.dimension_count" },
{ LLM_KV_ROPE_FREQ_BASE, "%s.rope.freq_base" },
{ LLM_KV_ROPE_SCALE_LINEAR, "%s.rope.scale_linear" },
{ LLM_KV_ROPE_SCALING_TYPE, "%s.rope.scaling.type" },
{ LLM_KV_ROPE_SCALING_FACTOR, "%s.rope.scaling.factor" },
{ LLM_KV_ROPE_SCALING_ORIG_CTX_LEN, "%s.rope.scaling.original_context_length" },
{ LLM_KV_ROPE_SCALING_FINETUNED, "%s.rope.scaling.finetuned" },
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
{ LLM_KV_SPLIT_NO, "split.no" },
{ LLM_KV_SPLIT_COUNT, "split.count" },
{ LLM_KV_SPLIT_TENSORS_COUNT, "split.tensors.count" },
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
{ LLM_KV_SSM_CONV_KERNEL, "%s.ssm.conv_kernel" },
{ LLM_KV_SSM_INNER_SIZE, "%s.ssm.inner_size" },
{ LLM_KV_SSM_STATE_SIZE, "%s.ssm.state_size" },
{ LLM_KV_SSM_TIME_STEP_RANK, "%s.ssm.time_step_rank" },
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
{ LLM_KV_TOKENIZER_MODEL, "tokenizer.ggml.model" },
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
{ LLM_KV_TOKENIZER_PRE, "tokenizer.ggml.pre" },
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
{ LLM_KV_TOKENIZER_LIST, "tokenizer.ggml.tokens" },
{ LLM_KV_TOKENIZER_TOKEN_TYPE, "tokenizer.ggml.token_type" },
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
{ LLM_KV_TOKENIZER_TOKEN_TYPE_COUNT, "tokenizer.ggml.token_type_count" },
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
{ LLM_KV_TOKENIZER_SCORES, "tokenizer.ggml.scores" },
{ LLM_KV_TOKENIZER_MERGES, "tokenizer.ggml.merges" },
{ LLM_KV_TOKENIZER_BOS_ID, "tokenizer.ggml.bos_token_id" },
{ LLM_KV_TOKENIZER_EOS_ID, "tokenizer.ggml.eos_token_id" },
{ LLM_KV_TOKENIZER_UNK_ID, "tokenizer.ggml.unknown_token_id" },
{ LLM_KV_TOKENIZER_SEP_ID, "tokenizer.ggml.seperator_token_id" },
{ LLM_KV_TOKENIZER_PAD_ID, "tokenizer.ggml.padding_token_id" },
BERT tokenizer fixes (#6498) Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
2024-04-09 17:44:08 +00:00
{ LLM_KV_TOKENIZER_CLS_ID, "tokenizer.ggml.cls_token_id" },
{ LLM_KV_TOKENIZER_MASK_ID, "tokenizer.ggml.mask_token_id" },
Respect tokenizer.ggml.add_bos_token value when tokenizing (#4040) * gguf-py: gguf-dump: Respect --no-tensor flag in JSON mode. * Respect add_bos_token GGUF metadata value * gguf-py: Try to fix SpecialVocab giving up too easily for the Nth time
2023-11-17 02:14:37 +00:00
{ LLM_KV_TOKENIZER_ADD_BOS, "tokenizer.ggml.add_bos_token" },
{ LLM_KV_TOKENIZER_ADD_EOS, "tokenizer.ggml.add_eos_token" },
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
{ LLM_KV_TOKENIZER_ADD_PREFIX, "tokenizer.ggml.add_space_prefix" },
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
{ LLM_KV_TOKENIZER_HF_JSON, "tokenizer.huggingface.json" },
{ LLM_KV_TOKENIZER_RWKV, "tokenizer.rwkv.world" },
gguf : add special tokens metadata for FIM/Infill (#6689) This commit adds special token metadata for Fill-In-the-Middle (FIM)/Infill to the GGUF model. The motivation for this is that currently there is support for CodeLlama but other models exist now like CodeGemma, but the different models use different token ids for the special tokens and this commit allows for supporting multiple models. Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
2024-04-16 06:13:13 +00:00
{ LLM_KV_TOKENIZER_PREFIX_ID, "tokenizer.ggml.prefix_token_id" },
{ LLM_KV_TOKENIZER_SUFFIX_ID, "tokenizer.ggml.suffix_token_id" },
{ LLM_KV_TOKENIZER_MIDDLE_ID, "tokenizer.ggml.middle_token_id" },
{ LLM_KV_TOKENIZER_EOT_ID, "tokenizer.ggml.eot_token_id" },
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
};
struct LLM_KV {
LLM_KV(llm_arch arch) : arch(arch) {}
llm_arch arch;
std::string operator()(llm_kv kv) const {
llama : fix segfault from unknown model arch name (#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion https://github.com/ggerganov/llama.cpp/pull/5820#issuecomment-1973735284 * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-03-02 13:42:56 +00:00
return ::format(LLM_KV_NAMES.at(kv), LLM_ARCH_NAMES.at(arch));
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
};
enum llm_tensor {
LLM_TENSOR_TOKEN_EMBD,
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
LLM_TENSOR_TOKEN_EMBD_NORM,
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
LLM_TENSOR_TOKEN_TYPES,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
LLM_TENSOR_POS_EMBD,
LLM_TENSOR_OUTPUT,
LLM_TENSOR_OUTPUT_NORM,
LLM_TENSOR_ROPE_FREQS,
LLM_TENSOR_ATTN_Q,
LLM_TENSOR_ATTN_K,
LLM_TENSOR_ATTN_V,
LLM_TENSOR_ATTN_QKV,
LLM_TENSOR_ATTN_OUT,
LLM_TENSOR_ATTN_NORM,
LLM_TENSOR_ATTN_NORM_2,
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
LLM_TENSOR_ATTN_OUT_NORM,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
LLM_TENSOR_ATTN_ROT_EMBD,
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
LLM_TENSOR_FFN_GATE_INP,
llama : add qwen2moe (#6074) * support qwen2moe * fix-review * metal : support unary ops for nelements % 4 != 0 * metal : require contiguousness for float4 unary kernels * metal : require contiguousness for float4 unary kernels (cont) * fix-review * names : for brevity "SHARED_EXP" -> "SHEXP" * llama : reuse build_moe_ffn() * llama : add model type name --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-16 15:40:48 +00:00
LLM_TENSOR_FFN_GATE_INP_SHEXP,
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
LLM_TENSOR_FFN_NORM,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
LLM_TENSOR_FFN_GATE,
LLM_TENSOR_FFN_DOWN,
LLM_TENSOR_FFN_UP,
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
LLM_TENSOR_FFN_ACT,
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
LLM_TENSOR_FFN_DOWN_EXP, // split experts for backward compatibility
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
LLM_TENSOR_FFN_GATE_EXP,
LLM_TENSOR_FFN_UP_EXP,
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
LLM_TENSOR_FFN_DOWN_EXPS, // merged experts
LLM_TENSOR_FFN_GATE_EXPS,
LLM_TENSOR_FFN_UP_EXPS,
llama : add qwen2moe (#6074) * support qwen2moe * fix-review * metal : support unary ops for nelements % 4 != 0 * metal : require contiguousness for float4 unary kernels * metal : require contiguousness for float4 unary kernels (cont) * fix-review * names : for brevity "SHARED_EXP" -> "SHEXP" * llama : reuse build_moe_ffn() * llama : add model type name --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-16 15:40:48 +00:00
LLM_TENSOR_FFN_DOWN_SHEXP,
LLM_TENSOR_FFN_GATE_SHEXP,
LLM_TENSOR_FFN_UP_SHEXP,
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
LLM_TENSOR_ATTN_Q_NORM,
LLM_TENSOR_ATTN_K_NORM,
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
LLM_TENSOR_LAYER_OUT_NORM,
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
LLM_TENSOR_SSM_IN,
LLM_TENSOR_SSM_CONV1D,
LLM_TENSOR_SSM_X,
LLM_TENSOR_SSM_DT,
LLM_TENSOR_SSM_A,
LLM_TENSOR_SSM_D,
LLM_TENSOR_SSM_OUT,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
};
llama : fix segfault from unknown model arch name (#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion https://github.com/ggerganov/llama.cpp/pull/5820#issuecomment-1973735284 * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-03-02 13:42:56 +00:00
static const std::map<llm_arch, std::map<llm_tensor, std::string>> LLM_TENSOR_NAMES = {
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
{
LLM_ARCH_LLAMA,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
{ LLM_TENSOR_FFN_GATE_EXP, "blk.%d.ffn_gate.%d" },
{ LLM_TENSOR_FFN_DOWN_EXP, "blk.%d.ffn_down.%d" },
{ LLM_TENSOR_FFN_UP_EXP, "blk.%d.ffn_up.%d" },
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
},
},
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
{
LLM_ARCH_BAICHUAN,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
{
LLM_ARCH_FALCON,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_NORM_2, "blk.%d.attn_norm_2" },
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
{
LLM_ARCH_GROK,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE_EXP, "blk.%d.ffn_gate.%d" },
{ LLM_TENSOR_FFN_DOWN_EXP, "blk.%d.ffn_down.%d" },
{ LLM_TENSOR_FFN_UP_EXP, "blk.%d.ffn_up.%d" },
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
{ LLM_TENSOR_LAYER_OUT_NORM, "blk.%d.layer_output_norm" },
{ LLM_TENSOR_ATTN_OUT_NORM, "blk.%d.attn_output_norm" },
},
},
examples : fix gpt-neox (#2943) Co-authored-by: mmnga <mmnga1mmnga@gmail.com>
2023-09-03 05:36:28 +00:00
{
LLM_ARCH_GPT2,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
{ LLM_TENSOR_POS_EMBD, "position_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
examples : fix gpt-neox (#2943) Co-authored-by: mmnga <mmnga1mmnga@gmail.com>
2023-09-03 05:36:28 +00:00
},
},
{
LLM_ARCH_GPTJ,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
},
},
{
LLM_ARCH_GPTNEOX,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
{
LLM_ARCH_PERSIMMON,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd"},
{ LLM_TENSOR_OUTPUT_NORM, "output_norm"},
{ LLM_TENSOR_OUTPUT, "output"},
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm"},
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv"},
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output"},
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm"},
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm"},
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm"},
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down"},
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up"},
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd"},
},
},
examples : fix gpt-neox (#2943) Co-authored-by: mmnga <mmnga1mmnga@gmail.com>
2023-09-03 05:36:28 +00:00
{
LLM_ARCH_MPT,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
mpt : implement backwards compatiblity with duped output tensor (#6139)
2024-03-18 16:49:02 +00:00
{ LLM_TENSOR_OUTPUT, "output"},
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
{ LLM_TENSOR_FFN_ACT, "blk.%d.ffn.act" },
llama : add SEA-LION support (#6448) * initial commit for sealion support * add sealion support * minor fix * q/k ln and pos_embd only if required * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * minor : clear whitespaces --------- Co-authored-by: bryan <bryansiow@aisingapore.org> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 18:05:10 +00:00
{ LLM_TENSOR_POS_EMBD, "position_embd" },
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm"},
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm"},
examples : fix gpt-neox (#2943) Co-authored-by: mmnga <mmnga1mmnga@gmail.com>
2023-09-03 05:36:28 +00:00
},
},
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
{
LLM_ARCH_STARCODER,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_POS_EMBD, "position_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
},
},
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
{
LLM_ARCH_REFACT,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
{
LLM_ARCH_BERT,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_TOKEN_EMBD_NORM, "token_embd_norm" },
{ LLM_TENSOR_TOKEN_TYPES, "token_types" },
{ LLM_TENSOR_POS_EMBD, "position_embd" },
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
{ LLM_TENSOR_ATTN_OUT_NORM, "blk.%d.attn_output_norm" },
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
{ LLM_TENSOR_LAYER_OUT_NORM, "blk.%d.layer_output_norm" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
{
LLM_ARCH_NOMIC_BERT,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_TOKEN_EMBD_NORM, "token_embd_norm" },
{ LLM_TENSOR_TOKEN_TYPES, "token_types" },
{ LLM_TENSOR_ATTN_OUT_NORM, "blk.%d.attn_output_norm" },
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_LAYER_OUT_NORM, "blk.%d.layer_output_norm" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
{
LLM_ARCH_BLOOM,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_TOKEN_EMBD_NORM, "token_embd_norm" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
},
},
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
{
LLM_ARCH_STABLELM,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
llama : add StableLM2 12B (#6635) * StableLM2 12B support for huggingface -> GGUF * StableLM12 tensormapping and constants * StableLM-2-12b model support * fix * Added 12B support * Removed autoformatting; resolved bug where model_arch was not selecting StableLM2 * Formatting * Do QK norm stacking in model conversion step * Converge StableLM and StableLM2 code to simplify graph construction * Fix accidental removal * Removed warnings * Revert formatter * Move QK norm stack to private function so it's easier to read * refactor stablelm graph builder to support 1.6, 3b and 12b more efficiently * Proper check for None type for new_name to avoid crash; formatting; revert change to base class `write_tensors()` * Format * Formatting * format Co-authored-by: compilade <git@compilade.net> * Fix incorrect check for K norm * space after commas; Keep indentation multiple of 4 spaces * Flake8 format * Removed unnecessary conditional branches * Removed unused comment * Fixed incorrect tensor passing * Format --------- Co-authored-by: compilade <git@compilade.net>
2024-04-16 15:48:35 +00:00
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
},
},
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
{
LLM_ARCH_QWEN,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
llama : support upcoming Qwen2 (#5037)
2024-01-19 11:53:13 +00:00
{
LLM_ARCH_QWEN2,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
llama : add qwen2moe (#6074) * support qwen2moe * fix-review * metal : support unary ops for nelements % 4 != 0 * metal : require contiguousness for float4 unary kernels * metal : require contiguousness for float4 unary kernels (cont) * fix-review * names : for brevity "SHARED_EXP" -> "SHEXP" * llama : reuse build_moe_ffn() * llama : add model type name --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-16 15:40:48 +00:00
{
LLM_ARCH_QWEN2MOE,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
{ LLM_TENSOR_FFN_GATE_INP_SHEXP, "blk.%d.ffn_gate_inp_shexp" },
{ LLM_TENSOR_FFN_GATE_SHEXP, "blk.%d.ffn_gate_shexp" },
{ LLM_TENSOR_FFN_DOWN_SHEXP, "blk.%d.ffn_down_shexp" },
{ LLM_TENSOR_FFN_UP_SHEXP, "blk.%d.ffn_up_shexp" },
},
},
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
{
LLM_ARCH_PHI2,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
convert : update phi-2 to latest HF repo (#4903) * convert : update phi-2 to latest HF repo ggml-ci * py : try to fix flake stuff
2024-01-13 11:44:37 +00:00
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
llama : add phi3 support (#6852) * add explicit phi3 support * add explicit phi3 support * remove unused code * convert : add BOS token * llama : match EOT token <|end|> * llama : minor / style * llama : tabs -> spaces * convert : fix lint checks --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-24 07:00:37 +00:00
{
LLM_ARCH_PHI3,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
{
LLM_ARCH_PLAMO,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
llama : add CodeShell support (#5016) * llama: add codeshell support * llama.cpp: fix codeshell with NeoX rope Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-19 09:07:27 +00:00
{
LLM_ARCH_CODESHELL,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
llama : add support for Orion-14B (#5118) * add support for Orion-14B(https://huggingface.co/OrionStarAI/Orion-14B-Chat) * flake8 support * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * Update llama.cpp * Update llama.cpp --------- Co-authored-by: lixiaopu <lixiaopu@cmcm.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-01-28 08:00:30 +00:00
{
LLM_ARCH_ORION,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
{
LLM_ARCH_INTERNLM2,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
{
LLM_ARCH_MINICPM,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
{ LLM_TENSOR_FFN_GATE_EXP, "blk.%d.ffn_gate.%d" },
{ LLM_TENSOR_FFN_DOWN_EXP, "blk.%d.ffn_down.%d" },
{ LLM_TENSOR_FFN_UP_EXP, "blk.%d.ffn_up.%d" },
},
},
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
{
LLM_ARCH_GEMMA,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
llama : add StarCoder2 support (#5795) * Add support for starcoder2 * handle rope type * skip rope freq and rotary embeddings from being serialized * resolve comments * Update llama.cpp * remove redundant changes * handle `rope-theta` * llama : change starcoder2 rope type * address comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-01 19:30:46 +00:00
{
LLM_ARCH_STARCODER2,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
{
LLM_ARCH_MAMBA,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_SSM_IN, "blk.%d.ssm_in" },
{ LLM_TENSOR_SSM_CONV1D, "blk.%d.ssm_conv1d" },
{ LLM_TENSOR_SSM_X, "blk.%d.ssm_x" },
{ LLM_TENSOR_SSM_DT, "blk.%d.ssm_dt" },
{ LLM_TENSOR_SSM_A, "blk.%d.ssm_a" },
{ LLM_TENSOR_SSM_D, "blk.%d.ssm_d" },
{ LLM_TENSOR_SSM_OUT, "blk.%d.ssm_out" },
},
},
[Model] Add support for xverse (#6301) * Support xverse model convert to gguf format. * 1. Convert xverse models to gguf; 2. Add LLM_ARCH_XVERSE inference in llama.cpp; 3. Add xverse item in Supported models in README.md; * * gguf-py: remove redundant logs * llama: remove the init_mapping_prefetch custom parameter * llama.cpp: Include the changes from #6122 to exclude the unused outputs of the last layers. * - Fix format issues - Remove duplicate set kqv_out to llm_build_kv * Update llama.cpp --------- Co-authored-by: willhe <willhe@xverse.cn> Co-authored-by: willhe <hexin@xverse.cn>
2024-03-29 13:37:03 +00:00
{
LLM_ARCH_XVERSE,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ROPE_FREQS, "rope_freqs" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_ATTN_ROT_EMBD, "blk.%d.attn_rot_embd" },
{ LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
{
LLM_ARCH_COMMAND_R,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
llama : add Command R Plus support (#6491) * Add Command R Plus GGUF * Add Command R Plus GGUF * Loading works up to LayerNorm2D * Export new tensors in 1D so they are not quantized. * Fix embedding layer based on Noeda's example * Whitespace * Add line * Fix unexpected tokens on MPS. Re-add F16 fix. ((Noeda) * dranger003: Fix block index overflow in CUDA dequantizing. * Reverted blocked multiplication code as it still has issues and could affect other Llama arches * export norms as f32 * fix overflow issues during quant and other cleanup * Type convention Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * dranger003: Fix more int overflow during quant. --------- Co-authored-by: S <seast@Ss-Mac-Studio.local> Co-authored-by: S <s@example.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-09 08:16:13 +00:00
{ LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" },
{ LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" },
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
},
},
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
{
LLM_ARCH_DBRX,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT_NORM, "output_norm" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" },
{ LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_ATTN_OUT_NORM, "blk.%d.attn_output_norm" },
{ LLM_TENSOR_FFN_GATE_INP, "blk.%d.ffn_gate_inp" },
{ LLM_TENSOR_FFN_GATE_EXPS, "blk.%d.ffn_gate_exps" },
{ LLM_TENSOR_FFN_DOWN_EXPS, "blk.%d.ffn_down_exps" },
{ LLM_TENSOR_FFN_UP_EXPS, "blk.%d.ffn_up_exps" },
},
},
Implement the OLMo architecture (#6741) * implement olmo architecture * remove unused variable * remove unused moe branch * remove check for weight * remove superfluous moe, bias and rope tensors * clarified comment * fix clamp_kqv setting * remove obsolete parameter name filter
2024-04-19 09:35:54 +00:00
{
LLM_ARCH_OLMO,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
{ LLM_TENSOR_OUTPUT, "output" },
{ LLM_TENSOR_ATTN_Q, "blk.%d.attn_q" },
{ LLM_TENSOR_ATTN_K, "blk.%d.attn_k" },
{ LLM_TENSOR_ATTN_V, "blk.%d.attn_v" },
{ LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" },
{ LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" },
{ LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" },
{ LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" },
},
},
examples : fix gpt-neox (#2943) Co-authored-by: mmnga <mmnga1mmnga@gmail.com>
2023-09-03 05:36:28 +00:00
{
LLM_ARCH_UNKNOWN,
{
{ LLM_TENSOR_TOKEN_EMBD, "token_embd" },
},
},
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
};
static llm_arch llm_arch_from_string(const std::string & name) {
for (const auto & kv : LLM_ARCH_NAMES) { // NOLINT
if (kv.second == name) {
return kv.first;
}
}
return LLM_ARCH_UNKNOWN;
}
// helper to handle gguf constants
// usage:
//
// const auto tn = LLM_TN(LLM_ARCH_LLAMA);
//
// std::string name = tn(LLM_TENSOR_OUTPUT); -> "output"
// std::string name = tn(LLM_TENSOR_TOKEN_EMBD, "bias"); -> "token_embd.bias"
// std::string name = tn(LLM_TENSOR_ATTN_NORM, "weight", 3); -> "blk.3.attn_norm.weight"
//
struct LLM_TN {
LLM_TN(llm_arch arch) : arch(arch) {}
llm_arch arch;
std::string operator()(llm_tensor tensor) const {
llama : fix segfault from unknown model arch name (#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion https://github.com/ggerganov/llama.cpp/pull/5820#issuecomment-1973735284 * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-03-02 13:42:56 +00:00
if (LLM_TENSOR_NAMES.at(arch).find(tensor) == LLM_TENSOR_NAMES.at(arch).end()) {
llama : fix quantization when tensors are missing (#5423)
2024-02-12 18:14:39 +00:00
return "__missing__";
}
llama : fix segfault from unknown model arch name (#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion https://github.com/ggerganov/llama.cpp/pull/5820#issuecomment-1973735284 * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-03-02 13:42:56 +00:00
return LLM_TENSOR_NAMES.at(arch).at(tensor);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
std::string operator()(llm_tensor tensor, const std::string & suffix) const {
llama : fix segfault from unknown model arch name (#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion https://github.com/ggerganov/llama.cpp/pull/5820#issuecomment-1973735284 * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-03-02 13:42:56 +00:00
if (LLM_TENSOR_NAMES.at(arch).find(tensor) == LLM_TENSOR_NAMES.at(arch).end()) {
llama : fix quantization when tensors are missing (#5423)
2024-02-12 18:14:39 +00:00
return "__missing__";
}
llama : fix segfault from unknown model arch name (#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion https://github.com/ggerganov/llama.cpp/pull/5820#issuecomment-1973735284 * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-03-02 13:42:56 +00:00
return LLM_TENSOR_NAMES.at(arch).at(tensor) + "." + suffix;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
std::string operator()(llm_tensor tensor, int bid) const {
llama : fix segfault from unknown model arch name (#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion https://github.com/ggerganov/llama.cpp/pull/5820#issuecomment-1973735284 * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-03-02 13:42:56 +00:00
if (LLM_TENSOR_NAMES.at(arch).find(tensor) == LLM_TENSOR_NAMES.at(arch).end()) {
llama : fix quantization when tensors are missing (#5423)
2024-02-12 18:14:39 +00:00
return "__missing__";
}
llama : fix segfault from unknown model arch name (#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion https://github.com/ggerganov/llama.cpp/pull/5820#issuecomment-1973735284 * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-03-02 13:42:56 +00:00
return ::format(LLM_TENSOR_NAMES.at(arch).at(tensor).c_str(), bid);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
std::string operator()(llm_tensor tensor, const std::string & suffix, int bid) const {
llama : fix segfault from unknown model arch name (#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion https://github.com/ggerganov/llama.cpp/pull/5820#issuecomment-1973735284 * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-03-02 13:42:56 +00:00
if (LLM_TENSOR_NAMES.at(arch).find(tensor) == LLM_TENSOR_NAMES.at(arch).end()) {
llama : fix quantization when tensors are missing (#5423)
2024-02-12 18:14:39 +00:00
return "__missing__";
}
llama : fix segfault from unknown model arch name (#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion https://github.com/ggerganov/llama.cpp/pull/5820#issuecomment-1973735284 * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-03-02 13:42:56 +00:00
return ::format(LLM_TENSOR_NAMES.at(arch).at(tensor).c_str(), bid) + "." + suffix;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
std::string operator()(llm_tensor tensor, const std::string & suffix, int bid, int xid) const {
llama : fix segfault from unknown model arch name (#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion https://github.com/ggerganov/llama.cpp/pull/5820#issuecomment-1973735284 * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-03-02 13:42:56 +00:00
if (LLM_TENSOR_NAMES.at(arch).find(tensor) == LLM_TENSOR_NAMES.at(arch).end()) {
llama : fix quantization when tensors are missing (#5423)
2024-02-12 18:14:39 +00:00
return "__missing__";
}
llama : fix segfault from unknown model arch name (#5820) * llama : fix segfault from unknown model arch name * llama : make all LLM maps const This also requires using `std::map::at` instead of its `operator[]` which does not exist for const maps. * llama : name LLM_ARCH_UNKNOWN to "(unknown)" This avoids errors from `std::map::at` when getting the general name of the model architecture. Using "(unknown)" instead of an empty string as per suggestion https://github.com/ggerganov/llama.cpp/pull/5820#issuecomment-1973735284 * llama : remove redundant inner const for LLM_TENSOR_NAMES The extra const won't do anything here as const maps return const references to values. Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * llama : remove redundant nullptr check in llm_arch_from_string Since LLM_ARCH_NAMES is a const map, no spurious elements with a NULL name are inserted anymore, so this check is dead code. --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-03-02 13:42:56 +00:00
return ::format(LLM_TENSOR_NAMES.at(arch).at(tensor).c_str(), bid, xid) + "." + suffix;
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
};
//
// gguf helpers
//
llama : allow for user specified embedding pooling type (#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-03 10:40:27 +00:00
static const std::map<llama_rope_scaling_type, const char *> LLAMA_ROPE_SCALING_TYPES = {
code : normalize enum names (#5697) * coda : normalize enum names ggml-ci * code : cont * code : cont
2024-02-25 10:09:09 +00:00
{ LLAMA_ROPE_SCALING_TYPE_NONE, "none" },
{ LLAMA_ROPE_SCALING_TYPE_LINEAR, "linear" },
{ LLAMA_ROPE_SCALING_TYPE_YARN, "yarn" },
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
};
llama : allow for user specified embedding pooling type (#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-03 10:40:27 +00:00
static llama_rope_scaling_type llama_rope_scaling_type_from_string(const std::string & name) {
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
for (const auto & kv : LLAMA_ROPE_SCALING_TYPES) {
if (kv.second == name) {
llama : allow for user specified embedding pooling type (#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-03 10:40:27 +00:00
return (llama_rope_scaling_type) kv.first;
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
}
}
code : normalize enum names (#5697) * coda : normalize enum names ggml-ci * code : cont * code : cont
2024-02-25 10:09:09 +00:00
return LLAMA_ROPE_SCALING_TYPE_UNSPECIFIED;
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
}
llama : add functions to get the model's metadata (#4013) * llama : add functions to get the model's metadata * format -> std::to_string * better documentation
2023-11-17 15:17:37 +00:00
static std::string gguf_data_to_str(enum gguf_type type, const void * data, int i) {
switch (type) {
case GGUF_TYPE_UINT8: return std::to_string(((const uint8_t *)data)[i]);
case GGUF_TYPE_INT8: return std::to_string(((const int8_t *)data)[i]);
case GGUF_TYPE_UINT16: return std::to_string(((const uint16_t *)data)[i]);
case GGUF_TYPE_INT16: return std::to_string(((const int16_t *)data)[i]);
case GGUF_TYPE_UINT32: return std::to_string(((const uint32_t *)data)[i]);
case GGUF_TYPE_INT32: return std::to_string(((const int32_t *)data)[i]);
case GGUF_TYPE_UINT64: return std::to_string(((const uint64_t *)data)[i]);
case GGUF_TYPE_INT64: return std::to_string(((const int64_t *)data)[i]);
case GGUF_TYPE_FLOAT32: return std::to_string(((const float *)data)[i]);
case GGUF_TYPE_FLOAT64: return std::to_string(((const double *)data)[i]);
case GGUF_TYPE_BOOL: return ((const bool *)data)[i] ? "true" : "false";
default: return format("unknown type %d", type);
}
}
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
static std::string gguf_kv_to_str(const struct gguf_context * ctx_gguf, int i) {
llama : add functions to get the model's metadata (#4013) * llama : add functions to get the model's metadata * format -> std::to_string * better documentation
2023-11-17 15:17:37 +00:00
const enum gguf_type type = gguf_get_kv_type(ctx_gguf, i);
switch (type) {
case GGUF_TYPE_STRING:
return gguf_get_val_str(ctx_gguf, i);
case GGUF_TYPE_ARRAY:
{
const enum gguf_type arr_type = gguf_get_arr_type(ctx_gguf, i);
int arr_n = gguf_get_arr_n(ctx_gguf, i);
const void * data = gguf_get_arr_data(ctx_gguf, i);
std::stringstream ss;
ss << "[";
for (int j = 0; j < arr_n; j++) {
if (arr_type == GGUF_TYPE_STRING) {
std::string val = gguf_get_arr_str(ctx_gguf, i, j);
// escape quotes
replace_all(val, "\\", "\\\\");
replace_all(val, "\"", "\\\"");
ss << '"' << val << '"';
} else if (arr_type == GGUF_TYPE_ARRAY) {
ss << "???";
} else {
ss << gguf_data_to_str(arr_type, data, j);
}
if (j < arr_n - 1) {
ss << ", ";
}
}
ss << "]";
return ss.str();
}
default:
return gguf_data_to_str(type, gguf_get_val_data(ctx_gguf, i), 0);
}
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
//
// llama helpers
//
#if defined(_WIN32)
static std::string llama_format_win_err(DWORD err) {
LPSTR buf;
size_t size = FormatMessageA(FORMAT_MESSAGE_ALLOCATE_BUFFER | FORMAT_MESSAGE_FROM_SYSTEM | FORMAT_MESSAGE_IGNORE_INSERTS,
NULL, err, MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT), (LPSTR)&buf, 0, NULL);
if (!size) {
return "FormatMessageA failed";
}
std::string ret(buf, size);
LocalFree(buf);
return ret;
}
#endif
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
template <typename T>
struct no_init {
T value;
no_init() { /* do nothing */ }
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
};
struct llama_file {
// use FILE * so we don't have to re-open the file to mmap
FILE * fp;
size_t size;
llama_file(const char * fname, const char * mode) {
use _wfopen instead of fopen on Windows (#6248) also fix missing #defines before windows.h, and BPE LF token on MSVC
2024-03-23 22:48:02 +00:00
fp = ggml_fopen(fname, mode);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
if (fp == NULL) {
throw std::runtime_error(format("failed to open %s: %s", fname, strerror(errno)));
}
seek(0, SEEK_END);
size = tell();
seek(0, SEEK_SET);
}
size_t tell() const {
#ifdef _WIN32
__int64 ret = _ftelli64(fp);
#else
long ret = std::ftell(fp);
#endif
GGML_ASSERT(ret != -1); // this really shouldn't fail
return (size_t) ret;
}
void seek(size_t offset, int whence) const {
#ifdef _WIN32
int ret = _fseeki64(fp, (__int64) offset, whence);
#else
int ret = std::fseek(fp, (long) offset, whence);
#endif
GGML_ASSERT(ret == 0); // same
}
void read_raw(void * ptr, size_t len) const {
if (len == 0) {
return;
}
errno = 0;
std::size_t ret = std::fread(ptr, len, 1, fp);
if (ferror(fp)) {
throw std::runtime_error(format("read error: %s", strerror(errno)));
}
if (ret != 1) {
llama : fix platforms without mmap (#4578) * llama : fix platforms without mmap * win32 : limit prefetch size to the file size * fix win32 error clobber, unnecessary std::string in std::runtime_error
2023-12-22 11:12:53 +00:00
throw std::runtime_error("unexpectedly reached end of file");
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
}
uint32_t read_u32() const {
uint32_t ret;
read_raw(&ret, sizeof(ret));
return ret;
}
void write_raw(const void * ptr, size_t len) const {
if (len == 0) {
return;
}
errno = 0;
size_t ret = std::fwrite(ptr, len, 1, fp);
if (ret != 1) {
throw std::runtime_error(format("write error: %s", strerror(errno)));
}
}
void write_u32(std::uint32_t val) const {
write_raw(&val, sizeof(val));
}
~llama_file() {
if (fp) {
std::fclose(fp);
}
}
};
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
using llama_files = std::vector<std::unique_ptr<llama_file>>;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
struct llama_mmap {
void * addr;
size_t size;
llama_mmap(const llama_mmap &) = delete;
#ifdef _POSIX_MAPPED_FILES
static constexpr bool SUPPORTED = true;
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
// list of mapped fragments (first_offset, last_offset)
std::vector<std::pair<size_t, size_t>> mapped_fragments;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama_mmap(struct llama_file * file, size_t prefetch = (size_t) -1 /* -1 = max value */, bool numa = false) {
size = file->size;
int fd = fileno(file->fp);
int flags = MAP_SHARED;
// prefetch/readahead impairs performance on NUMA systems
ggml : add numa options (#5377) * Added numa options to allow finer grained control as well as plumbing for a new mirror mode that will require numa.h * Reverted Makefile * Fixed include * Removed sched.h from ggml.h, moved ggml_get_numa_affinity into ggml.c, removed trailing whitespace and fixed up a few inconsistent variables * removed trailing whitespace * Added numa options to allow finer grained control as well as plumbing for a new mirror mode that will require numa.h * Reverting Makefile * Fixed a number of issues with the move from BOOL to ggml_numa_strategies. Added a note about mirror mode note being implemented yet * Removing MIRROR_MODE code for this PR * Removing last bit of MIRROR_MODE code for this PR * Removing unneeded branch in server.cpp example and moving get_numa_affinity and making it static * Fixed lingering init_llama_backend() bool calls in tests and examples * Remote enum llama_numa_strategies * Revert bad merge with dynatemp flags * add missing enum ggml_numa_strategies declaration and revert sync problem with master * add missing enum ggml_numa_strategies declaration * fixed ggml_init_numa variable * Update ggml.h Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Update READMEs with info about numa flags, change INTERLEAVE strategy name to DISTRIBUTE everywhere, implement the improved distribution strategy from @rankaiyx, fix a spelling mistake and un-merge some bad merges * split numa init out from llama_backend_init and created llama_numa_init. Updated all code paths and samples * Fix up some boolean vs enum comparisons * Added #ifdefs for non-Linux OS that don't have cpu_set_t datatype * Update ggml.h Align enum values Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml.c Remove whitespace Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml.c align paremeters Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update examples/server/server.cpp remove whitespace and align brace Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update common/common.cpp Remove whitespace and align brace Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * unified ggml_numa_strategy enum and fixed text alignment in server.cpp example * Update ggml.c simplified return for platforms without NUMA support Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * removed redundant else from cli argument processing of --numa * whitespace --------- Co-authored-by: root <root@nenya.lothlorien.ca> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Jared Van Bortel <jared@nomic.ai>
2024-02-16 09:31:07 +00:00
if (numa) { prefetch = 0; }
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#ifdef __linux__
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
// advise the kernel to read the file sequentially (increases readahead)
if (posix_fadvise(fd, 0, 0, POSIX_FADV_SEQUENTIAL)) {
LLAMA_LOG_WARN("warning: posix_fadvise(.., POSIX_FADV_SEQUENTIAL) failed: %s\n",
strerror(errno));
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
if (prefetch) { flags |= MAP_POPULATE; }
#endif
addr = mmap(NULL, file->size, PROT_READ, flags, fd, 0);
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
if (addr == MAP_FAILED) { // NOLINT
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
throw std::runtime_error(format("mmap failed: %s", strerror(errno)));
}
if (prefetch > 0) {
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
// advise the kernel to preload the mapped memory
ggml : posixify madvise and pagesize (#3037) * llama : use posix_madvise() instead of madvise() derived from BSD sed -i 's,\<madvise\>,posix_&,g;s,\<MADV_,POSIX_&,g' llama.cpp * ggml : use sysconf(_SC_PAGESIZE) instead of getpagesize() derived from BSD sed -i 's,getpagesize(),sysconf(_SC_PAGESIZE),g' ggml.c * metal : use sysconf(_SC_PAGESIZE) instead of getpagesize() derived from BSD sed -i 's,getpagesize(),sysconf(_SC_PAGESIZE),g' ggml-metal.m
2023-09-07 08:15:06 +00:00
if (posix_madvise(addr, std::min(file->size, prefetch), POSIX_MADV_WILLNEED)) {
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
LLAMA_LOG_WARN("warning: posix_madvise(.., POSIX_MADV_WILLNEED) failed: %s\n",
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
strerror(errno));
}
}
if (numa) {
// advise the kernel not to use readahead
// (because the next page might not belong on the same node)
ggml : posixify madvise and pagesize (#3037) * llama : use posix_madvise() instead of madvise() derived from BSD sed -i 's,\<madvise\>,posix_&,g;s,\<MADV_,POSIX_&,g' llama.cpp * ggml : use sysconf(_SC_PAGESIZE) instead of getpagesize() derived from BSD sed -i 's,getpagesize(),sysconf(_SC_PAGESIZE),g' ggml.c * metal : use sysconf(_SC_PAGESIZE) instead of getpagesize() derived from BSD sed -i 's,getpagesize(),sysconf(_SC_PAGESIZE),g' ggml-metal.m
2023-09-07 08:15:06 +00:00
if (posix_madvise(addr, file->size, POSIX_MADV_RANDOM)) {
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
LLAMA_LOG_WARN("warning: posix_madvise(.., POSIX_MADV_RANDOM) failed: %s\n",
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
strerror(errno));
}
}
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
// initialize list of mapped_fragments
mapped_fragments.emplace_back(0, file->size);
}
static void align_range(size_t * first, size_t * last, size_t page_size) {
// align first to the next page
size_t offset_in_page = *first & (page_size - 1);
size_t offset_to_page = offset_in_page == 0 ? 0 : page_size - offset_in_page;
*first += offset_to_page;
// align last to the previous page
*last = *last & ~(page_size - 1);
if (*last <= *first) {
*last = *first;
}
}
// partially unmap the file in the range [first, last)
void unmap_fragment(size_t first, size_t last) {
// note: this function must not be called multiple times with overlapping ranges
// otherwise, there is a risk of invalidating addresses that have been repurposed for other mappings
int page_size = sysconf(_SC_PAGESIZE);
align_range(&first, &last, page_size);
size_t len = last - first;
if (len == 0) {
return;
}
GGML_ASSERT(first % page_size == 0);
GGML_ASSERT(last % page_size == 0);
GGML_ASSERT(last > first);
void * next_page_start = (uint8_t *) addr + first;
// unmap the range
if (munmap(next_page_start, len)) {
LLAMA_LOG_WARN("warning: munmap failed: %s\n", strerror(errno));
}
// update the list of mapped fragments to avoid unmapping the same range again in the destructor
std::vector<std::pair<size_t, size_t>> new_mapped_fragments;
for (const auto & frag : mapped_fragments) {
if (frag.first < first && frag.second > last) {
// the range is in the middle of the fragment, split it
new_mapped_fragments.emplace_back(frag.first, first);
new_mapped_fragments.emplace_back(last, frag.second);
} else if (frag.first < first && frag.second > first) {
// the range starts in the middle of the fragment
new_mapped_fragments.emplace_back(frag.first, first);
} else if (frag.first < last && frag.second > last) {
// the range ends in the middle of the fragment
new_mapped_fragments.emplace_back(last, frag.second);
} else if (frag.first >= first && frag.second <= last) {
// the range covers the entire fragment
} else {
// the range is outside the fragment
new_mapped_fragments.push_back(frag);
}
}
mapped_fragments = std::move(new_mapped_fragments);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
~llama_mmap() {
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
for (const auto & frag : mapped_fragments) {
if (munmap((char *) addr + frag.first, frag.second - frag.first)) {
LLAMA_LOG_WARN("warning: munmap failed: %s\n", strerror(errno));
}
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
#elif defined(_WIN32)
static constexpr bool SUPPORTED = true;
llama : fix platforms without mmap (#4578) * llama : fix platforms without mmap * win32 : limit prefetch size to the file size * fix win32 error clobber, unnecessary std::string in std::runtime_error
2023-12-22 11:12:53 +00:00
llama_mmap(struct llama_file * file, size_t prefetch = (size_t) -1, bool numa = false) {
GGML_UNUSED(numa);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
size = file->size;
HANDLE hFile = (HANDLE) _get_osfhandle(_fileno(file->fp));
HANDLE hMapping = CreateFileMappingA(hFile, NULL, PAGE_READONLY, 0, 0, NULL);
if (hMapping == NULL) {
llama : fix platforms without mmap (#4578) * llama : fix platforms without mmap * win32 : limit prefetch size to the file size * fix win32 error clobber, unnecessary std::string in std::runtime_error
2023-12-22 11:12:53 +00:00
DWORD error = GetLastError();
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
throw std::runtime_error(format("CreateFileMappingA failed: %s", llama_format_win_err(error).c_str()));
}
addr = MapViewOfFile(hMapping, FILE_MAP_READ, 0, 0, 0);
llama : fix platforms without mmap (#4578) * llama : fix platforms without mmap * win32 : limit prefetch size to the file size * fix win32 error clobber, unnecessary std::string in std::runtime_error
2023-12-22 11:12:53 +00:00
DWORD error = GetLastError();
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
CloseHandle(hMapping);
if (addr == NULL) {
throw std::runtime_error(format("MapViewOfFile failed: %s", llama_format_win_err(error).c_str()));
}
llama : fix platforms without mmap (#4578) * llama : fix platforms without mmap * win32 : limit prefetch size to the file size * fix win32 error clobber, unnecessary std::string in std::runtime_error
2023-12-22 11:12:53 +00:00
if (prefetch > 0) {
llama : support WinXP build with MinGW 8.1.0 (#3419)
2024-01-14 08:41:44 +00:00
#if _WIN32_WINNT >= 0x602
@vxiiduu's fix for PrefetchVirtualMemory (#2930) Reimplement fix for `PrefetchVirtualMemory`. Co-authored-by: vxiiduu <73044267+vxiiduu@users.noreply.github.com>
2023-08-31 11:21:45 +00:00
// PrefetchVirtualMemory is only present on Windows 8 and above, so we dynamically load it
BOOL (WINAPI *pPrefetchVirtualMemory) (HANDLE, ULONG_PTR, PWIN32_MEMORY_RANGE_ENTRY, ULONG);
HMODULE hKernel32 = GetModuleHandleW(L"kernel32.dll");
// may fail on pre-Windows 8 systems
pPrefetchVirtualMemory = reinterpret_cast<decltype(pPrefetchVirtualMemory)> (GetProcAddress(hKernel32, "PrefetchVirtualMemory"));
if (pPrefetchVirtualMemory) {
// advise the kernel to preload the mapped memory
WIN32_MEMORY_RANGE_ENTRY range;
range.VirtualAddress = addr;
llama : fix platforms without mmap (#4578) * llama : fix platforms without mmap * win32 : limit prefetch size to the file size * fix win32 error clobber, unnecessary std::string in std::runtime_error
2023-12-22 11:12:53 +00:00
range.NumberOfBytes = (SIZE_T) std::min(size, prefetch);
@vxiiduu's fix for PrefetchVirtualMemory (#2930) Reimplement fix for `PrefetchVirtualMemory`. Co-authored-by: vxiiduu <73044267+vxiiduu@users.noreply.github.com>
2023-08-31 11:21:45 +00:00
if (!pPrefetchVirtualMemory(GetCurrentProcess(), 1, &range, 0)) {
llama : fix platforms without mmap (#4578) * llama : fix platforms without mmap * win32 : limit prefetch size to the file size * fix win32 error clobber, unnecessary std::string in std::runtime_error
2023-12-22 11:12:53 +00:00
LLAMA_LOG_WARN("warning: PrefetchVirtualMemory failed: %s\n",
@vxiiduu's fix for PrefetchVirtualMemory (#2930) Reimplement fix for `PrefetchVirtualMemory`. Co-authored-by: vxiiduu <73044267+vxiiduu@users.noreply.github.com>
2023-08-31 11:21:45 +00:00
llama_format_win_err(GetLastError()).c_str());
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llama : support WinXP build with MinGW 8.1.0 (#3419)
2024-01-14 08:41:44 +00:00
#else
throw std::runtime_error("PrefetchVirtualMemory unavailable");
#endif
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
}
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
void unmap_fragment(size_t first, size_t last) {
// not supported
GGML_UNUSED(first);
GGML_UNUSED(last);
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
~llama_mmap() {
if (!UnmapViewOfFile(addr)) {
llama : fix platforms without mmap (#4578) * llama : fix platforms without mmap * win32 : limit prefetch size to the file size * fix win32 error clobber, unnecessary std::string in std::runtime_error
2023-12-22 11:12:53 +00:00
LLAMA_LOG_WARN("warning: UnmapViewOfFile failed: %s\n",
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama_format_win_err(GetLastError()).c_str());
}
}
#else
static constexpr bool SUPPORTED = false;
llama : fix platforms without mmap (#4578) * llama : fix platforms without mmap * win32 : limit prefetch size to the file size * fix win32 error clobber, unnecessary std::string in std::runtime_error
2023-12-22 11:12:53 +00:00
llama_mmap(struct llama_file * file, size_t prefetch = -1, bool numa = false) {
GGML_UNUSED(file);
GGML_UNUSED(prefetch);
GGML_UNUSED(numa);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : fix platforms without mmap (#4578) * llama : fix platforms without mmap * win32 : limit prefetch size to the file size * fix win32 error clobber, unnecessary std::string in std::runtime_error
2023-12-22 11:12:53 +00:00
throw std::runtime_error("mmap not supported");
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
llama : fix platforms without mmap (#4578) * llama : fix platforms without mmap * win32 : limit prefetch size to the file size * fix win32 error clobber, unnecessary std::string in std::runtime_error
2023-12-22 11:12:53 +00:00
void unmap_fragment(size_t first, size_t last) {
GGML_UNUSED(first);
GGML_UNUSED(last);
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
llama : fix platforms without mmap (#4578) * llama : fix platforms without mmap * win32 : limit prefetch size to the file size * fix win32 error clobber, unnecessary std::string in std::runtime_error
2023-12-22 11:12:53 +00:00
throw std::runtime_error("mmap not supported");
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#endif
};
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
using llama_mmaps = std::vector<std::unique_ptr<llama_mmap>>;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
// Represents some region of memory being locked using mlock or VirtualLock;
// will automatically unlock on destruction.
struct llama_mlock {
void * addr = NULL;
size_t size = 0;
bool failed_already = false;
llama_mlock() {}
llama_mlock(const llama_mlock &) = delete;
~llama_mlock() {
if (size) {
raw_unlock(addr, size);
}
}
void init(void * ptr) {
GGML_ASSERT(addr == NULL && size == 0); // NOLINT
addr = ptr;
}
void grow_to(size_t target_size) {
GGML_ASSERT(addr);
if (failed_already) {
return;
}
size_t granularity = lock_granularity();
target_size = (target_size + granularity - 1) & ~(granularity - 1);
if (target_size > size) {
if (raw_lock((uint8_t *) addr + size, target_size - size)) {
size = target_size;
} else {
failed_already = true;
}
}
}
#ifdef _POSIX_MEMLOCK_RANGE
static constexpr bool SUPPORTED = true;
static size_t lock_granularity() {
return (size_t) sysconf(_SC_PAGESIZE);
}
#ifdef __APPLE__
#define MLOCK_SUGGESTION \
"Try increasing the sysctl values 'vm.user_wire_limit' and 'vm.global_user_wire_limit' and/or " \
fix typo "RLIMIT_MLOCK" (#5175)
2024-01-29 14:45:41 +00:00
"decreasing 'vm.global_no_user_wire_amount'. Also try increasing RLIMIT_MEMLOCK (ulimit -l).\n"
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#else
#define MLOCK_SUGGESTION \
fix typo "RLIMIT_MLOCK" (#5175)
2024-01-29 14:45:41 +00:00
"Try increasing RLIMIT_MEMLOCK ('ulimit -l' as root).\n"
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#endif
bool raw_lock(const void * addr, size_t size) const {
if (!mlock(addr, size)) {
return true;
}
char* errmsg = std::strerror(errno);
bool suggest = (errno == ENOMEM);
// Check if the resource limit is fine after all
struct rlimit lock_limit;
if (suggest && getrlimit(RLIMIT_MEMLOCK, &lock_limit)) {
suggest = false;
}
if (suggest && (lock_limit.rlim_max > lock_limit.rlim_cur + size)) {
suggest = false;
}
llama : use LLAMA_LOG_ macros for logging
2024-01-14 09:03:19 +00:00
LLAMA_LOG_WARN("warning: failed to mlock %zu-byte buffer (after previously locking %zu bytes): %s\n%s",
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
size, this->size, errmsg, suggest ? MLOCK_SUGGESTION : "");
return false;
}
#undef MLOCK_SUGGESTION
static void raw_unlock(void * addr, size_t size) {
if (munlock(addr, size)) {
llama : use LLAMA_LOG_ macros for logging
2024-01-14 09:03:19 +00:00
LLAMA_LOG_WARN("warning: failed to munlock buffer: %s\n", std::strerror(errno));
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
}
#elif defined(_WIN32)
static constexpr bool SUPPORTED = true;
static size_t lock_granularity() {
SYSTEM_INFO si;
GetSystemInfo(&si);
return (size_t) si.dwPageSize;
}
bool raw_lock(void * ptr, size_t len) const {
for (int tries = 1; ; tries++) {
if (VirtualLock(ptr, len)) {
return true;
}
if (tries == 2) {
llama : use LLAMA_LOG_ macros for logging
2024-01-14 09:03:19 +00:00
LLAMA_LOG_WARN("warning: failed to VirtualLock %zu-byte buffer (after previously locking %zu bytes): %s\n",
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
len, size, llama_format_win_err(GetLastError()).c_str());
return false;
}
// It failed but this was only the first try; increase the working
// set size and try again.
SIZE_T min_ws_size, max_ws_size;
if (!GetProcessWorkingSetSize(GetCurrentProcess(), &min_ws_size, &max_ws_size)) {
llama : use LLAMA_LOG_ macros for logging
2024-01-14 09:03:19 +00:00
LLAMA_LOG_WARN("warning: GetProcessWorkingSetSize failed: %s\n",
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama_format_win_err(GetLastError()).c_str());
return false;
}
// Per MSDN: "The maximum number of pages that a process can lock
// is equal to the number of pages in its minimum working set minus
// a small overhead."
// Hopefully a megabyte is enough overhead:
size_t increment = len + 1048576;
// The minimum must be <= the maximum, so we need to increase both:
min_ws_size += increment;
max_ws_size += increment;
if (!SetProcessWorkingSetSize(GetCurrentProcess(), min_ws_size, max_ws_size)) {
llama : use LLAMA_LOG_ macros for logging
2024-01-14 09:03:19 +00:00
LLAMA_LOG_WARN("warning: SetProcessWorkingSetSize failed: %s\n",
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama_format_win_err(GetLastError()).c_str());
return false;
}
}
}
static void raw_unlock(void * ptr, size_t len) {
if (!VirtualUnlock(ptr, len)) {
llama : use LLAMA_LOG_ macros for logging
2024-01-14 09:03:19 +00:00
LLAMA_LOG_WARN("warning: failed to VirtualUnlock buffer: %s\n",
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama_format_win_err(GetLastError()).c_str());
}
}
#else
static constexpr bool SUPPORTED = false;
static size_t lock_granularity() {
return (size_t) 65536;
}
bool raw_lock(const void * addr, size_t len) const {
llama : use LLAMA_LOG_ macros for logging
2024-01-14 09:03:19 +00:00
LLAMA_LOG_WARN("warning: mlock not supported on this system\n");
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
return false;
}
static void raw_unlock(const void * addr, size_t len) {}
#endif
};
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
using llama_mlocks = std::vector<std::unique_ptr<llama_mlock>>;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : add option to render special/control tokens (#6807) * make : fix common dep on llama.h * llama : add option to render special tokens * readme : add API change notice ggml-ci * swift : fix build
2024-04-21 15:36:45 +00:00
static std::string llama_token_to_piece(const struct llama_context * ctx, llama_token token, bool special) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
std::vector<char> result(8, 0);
llama : add option to render special/control tokens (#6807) * make : fix common dep on llama.h * llama : add option to render special tokens * readme : add API change notice ggml-ci * swift : fix build
2024-04-21 15:36:45 +00:00
const int n_tokens = llama_token_to_piece(llama_get_model(ctx), token, result.data(), result.size(), special);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
if (n_tokens < 0) {
result.resize(-n_tokens);
llama : add option to render special/control tokens (#6807) * make : fix common dep on llama.h * llama : add option to render special tokens * readme : add API change notice ggml-ci * swift : fix build
2024-04-21 15:36:45 +00:00
int check = llama_token_to_piece(llama_get_model(ctx), token, result.data(), result.size(), special);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
GGML_ASSERT(check == -n_tokens);
Add test for MPT tokenization (#3728) * Add test for MPT tokenization * Revert code motion * Remove unnecessary restriction in test case * Clarify logic in conversion
2023-10-22 19:21:42 +00:00
}
else {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
result.resize(n_tokens);
}
return std::string(result.data(), result.size());
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
static ggml_backend_buffer_type_t llama_default_buffer_type_cpu(bool host_buffer) {
fallback to CPU buffer if host buffer alloc fails (#4610)
2023-12-23 15:10:51 +00:00
ggml_backend_buffer_type_t buft = nullptr;
cuda : rename build flag to LLAMA_CUDA (#6299)
2024-03-26 00:16:01 +00:00
#if defined(GGML_USE_CUDA)
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// host buffers should only be used when data is expected to be copied to/from the GPU
if (host_buffer) {
buft = ggml_backend_cuda_host_buffer_type();
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
ggml : add unified SYCL backend for Intel GPUs (#2690) * first update for migration * update init_cublas * add debug functio, commit all help code * step 1 * step 2 * step3 add fp16, slower 31->28 * add GGML_LIST_DEVICE function * step 5 format device and print * step6, enhance error check, remove CUDA macro, enhance device id to fix none-zero id issue * support main device is non-zero * step7 add debug for code path, rm log * step 8, rename all macro & func from cuda by sycl * fix error of select non-zero device, format device list * ren ggml-sycl.hpp -> ggml-sycl.h * clear CMAKE to rm unused lib and options * correct queue: rm dtct:get_queue * add print tensor function to debug * fix error: wrong result in 658746bb26702e50f2c59c0e4ada8e9da6010481 * summary dpct definition in one header file to replace folder:dpct * refactor device log * mv dpct definition from folder dpct to ggml-sycl.h * update readme, refactor build script * fix build with sycl * set nthread=1 when sycl, increase performance * add run script, comment debug code * add ls-sycl-device tool * add ls-sycl-device, rm unused files * rm rear space * dos2unix * Update README_sycl.md * fix return type * remove sycl version from include path * restore rm code to fix hang issue * add syc and link for sycl readme * rm original sycl code before refactor * fix code err * add know issue for pvc hang issue * enable SYCL_F16 support * align pr4766 * check for sycl blas, better performance * cleanup 1 * remove extra endif * add build&run script, clean CMakefile, update guide by review comments * rename macro to intel hardware * editor config format * format fixes * format fixes * editor format fix * Remove unused headers * skip build sycl tool for other code path * replace tab by space * fix blas matmul function * fix mac build * restore hip dependency * fix conflict * ren as review comments * mv internal function to .cpp file * export funciton print_sycl_devices(), mv class dpct definition to source file * update CI/action for sycl code, fix CI error of repeat/dup * fix action ID format issue * rm unused strategy * enable llama_f16 in ci * fix conflict * fix build break on MacOS, due to CI of MacOS depend on external ggml, instead of internal ggml * fix ci cases for unsupported data type * revert unrelated changed in cuda cmake remove useless nommq fix typo of GGML_USE_CLBLAS_SYCL * revert hip cmake changes * fix indent * add prefix in func name * revert no mmq * rm cpu blas duplicate * fix no_new_line * fix src1->type==F16 bug. * pass batch offset for F16 src1 * fix batch error * fix wrong code * revert sycl checking in test-sampling * pass void as arguments of ggml_backend_sycl_print_sycl_devices * remove extra blank line in test-sampling * revert setting n_threads in sycl * implement std::isinf for icpx with fast math. * Update ci/run.sh Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update examples/sycl/run-llama2.sh Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update examples/sycl/run-llama2.sh Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * add copyright and MIT license declare * update the cmd example --------- Co-authored-by: jianyuzh <jianyu.zhang@intel.com> Co-authored-by: luoyu-intel <yu.luo@intel.com> Co-authored-by: Meng, Hengyu <hengyu.meng@intel.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 15:56:23 +00:00
#elif defined(GGML_USE_SYCL)
Support multiple GPUs (split mode) on SYCL backend (#5806) * suport multiple cards: split-mode - layer|row * rm warning * rebase with master, support tow new OPs, close feature for -sm=row, fix for unit test * update news * fix merge error * update according to review comments
2024-03-02 11:49:30 +00:00
if (host_buffer) {
buft = ggml_backend_sycl_host_buffer_type();
}
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
#elif defined(GGML_USE_CPU_HBM)
fallback to CPU buffer if host buffer alloc fails (#4610)
2023-12-23 15:10:51 +00:00
buft = ggml_backend_cpu_hbm_buffer_type();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 17:03:59 +00:00
#elif defined(GGML_USE_VULKAN)
if (host_buffer) {
buft = ggml_backend_vk_host_buffer_type();
}
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
#endif
fallback to CPU buffer if host buffer alloc fails (#4610)
2023-12-23 15:10:51 +00:00
if (buft == nullptr) {
buft = ggml_backend_cpu_buffer_type();
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
return buft;
GGML_UNUSED(host_buffer);
}
static ggml_backend_buffer_type_t llama_default_buffer_type_offload(int gpu) {
ggml_backend_buffer_type_t buft = nullptr;
#ifdef GGML_USE_METAL
buft = ggml_backend_metal_buffer_type();
cuda : rename build flag to LLAMA_CUDA (#6299)
2024-03-26 00:16:01 +00:00
#elif defined(GGML_USE_CUDA)
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
buft = ggml_backend_cuda_buffer_type(gpu);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 17:03:59 +00:00
#elif defined(GGML_USE_VULKAN)
Basic Vulkan Multi-GPU implementation (#5321) * Initial Vulkan multi-gpu implementation Move most global variables into backend context * Add names to backend device functions * Add further missing cleanup code * Reduce code duplication in tensor split layer assignment * generalize LLAMA_SPLIT_LAYER for all backends, do not expose device count and memory in llama.h * Only do device info print in the beginning and initialize one backend for cpu assist Add missing cleanup code * Rework backend memory management to make sure devices and buffers get properly allocated and freed * Rename cpu assist free function --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-07 06:54:50 +00:00
buft = ggml_backend_vk_buffer_type(gpu);
ggml : add unified SYCL backend for Intel GPUs (#2690) * first update for migration * update init_cublas * add debug functio, commit all help code * step 1 * step 2 * step3 add fp16, slower 31->28 * add GGML_LIST_DEVICE function * step 5 format device and print * step6, enhance error check, remove CUDA macro, enhance device id to fix none-zero id issue * support main device is non-zero * step7 add debug for code path, rm log * step 8, rename all macro & func from cuda by sycl * fix error of select non-zero device, format device list * ren ggml-sycl.hpp -> ggml-sycl.h * clear CMAKE to rm unused lib and options * correct queue: rm dtct:get_queue * add print tensor function to debug * fix error: wrong result in 658746bb26702e50f2c59c0e4ada8e9da6010481 * summary dpct definition in one header file to replace folder:dpct * refactor device log * mv dpct definition from folder dpct to ggml-sycl.h * update readme, refactor build script * fix build with sycl * set nthread=1 when sycl, increase performance * add run script, comment debug code * add ls-sycl-device tool * add ls-sycl-device, rm unused files * rm rear space * dos2unix * Update README_sycl.md * fix return type * remove sycl version from include path * restore rm code to fix hang issue * add syc and link for sycl readme * rm original sycl code before refactor * fix code err * add know issue for pvc hang issue * enable SYCL_F16 support * align pr4766 * check for sycl blas, better performance * cleanup 1 * remove extra endif * add build&run script, clean CMakefile, update guide by review comments * rename macro to intel hardware * editor config format * format fixes * format fixes * editor format fix * Remove unused headers * skip build sycl tool for other code path * replace tab by space * fix blas matmul function * fix mac build * restore hip dependency * fix conflict * ren as review comments * mv internal function to .cpp file * export funciton print_sycl_devices(), mv class dpct definition to source file * update CI/action for sycl code, fix CI error of repeat/dup * fix action ID format issue * rm unused strategy * enable llama_f16 in ci * fix conflict * fix build break on MacOS, due to CI of MacOS depend on external ggml, instead of internal ggml * fix ci cases for unsupported data type * revert unrelated changed in cuda cmake remove useless nommq fix typo of GGML_USE_CLBLAS_SYCL * revert hip cmake changes * fix indent * add prefix in func name * revert no mmq * rm cpu blas duplicate * fix no_new_line * fix src1->type==F16 bug. * pass batch offset for F16 src1 * fix batch error * fix wrong code * revert sycl checking in test-sampling * pass void as arguments of ggml_backend_sycl_print_sycl_devices * remove extra blank line in test-sampling * revert setting n_threads in sycl * implement std::isinf for icpx with fast math. * Update ci/run.sh Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update examples/sycl/run-llama2.sh Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update examples/sycl/run-llama2.sh Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * add copyright and MIT license declare * update the cmd example --------- Co-authored-by: jianyuzh <jianyu.zhang@intel.com> Co-authored-by: luoyu-intel <yu.luo@intel.com> Co-authored-by: Meng, Hengyu <hengyu.meng@intel.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 15:56:23 +00:00
#elif defined(GGML_USE_SYCL)
buft = ggml_backend_sycl_buffer_type(gpu);
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
#elif defined(GGML_USE_CLBLAST)
buft = ggml_backend_opencl_buffer_type();
Nomic Vulkan backend (#4456) Signed-off-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: niansa <anton-sa@web.de> Co-authored-by: Adam Treat <treat.adam@gmail.com> Co-authored-by: Aaron Miller <apage43@ninjawhale.com> Co-authored-by: ToKiNoBug <tokinobug@163.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-01-29 20:50:50 +00:00
#elif defined(GGML_USE_KOMPUTE)
buft = ggml_backend_kompute_buffer_type(gpu);
if (buft == nullptr) {
LLAMA_LOG_WARN("%s: cannot use GPU %d, check `vulkaninfo --summary`\n", __func__, gpu);
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
#endif
if (buft == nullptr) {
buft = llama_default_buffer_type_cpu(true);
}
return buft;
GGML_UNUSED(gpu);
}
static ggml_backend_buffer_type_t llama_default_buffer_type_split(int fallback_gpu, const float * tensor_split) {
ggml_backend_buffer_type_t buft = nullptr;
cuda : rename build flag to LLAMA_CUDA (#6299)
2024-03-26 00:16:01 +00:00
#ifdef GGML_USE_CUDA
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
if (ggml_backend_cuda_get_device_count() > 1) {
buft = ggml_backend_cuda_split_buffer_type(tensor_split);
}
#endif
fallback to CPU buffer if host buffer alloc fails (#4610)
2023-12-23 15:10:51 +00:00
Support multiple GPUs (split mode) on SYCL backend (#5806) * suport multiple cards: split-mode - layer|row * rm warning * rebase with master, support tow new OPs, close feature for -sm=row, fix for unit test * update news * fix merge error * update according to review comments
2024-03-02 11:49:30 +00:00
#ifdef GGML_USE_SYCL
if (ggml_backend_sycl_get_device_count() > 1) {
buft = ggml_backend_sycl_split_buffer_type(tensor_split);
}
#endif
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
if (buft == nullptr) {
buft = llama_default_buffer_type_offload(fallback_gpu);
}
fallback to CPU buffer if host buffer alloc fails (#4610)
2023-12-23 15:10:51 +00:00
return buft;
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
GGML_UNUSED(tensor_split);
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
Basic Vulkan Multi-GPU implementation (#5321) * Initial Vulkan multi-gpu implementation Move most global variables into backend context * Add names to backend device functions * Add further missing cleanup code * Reduce code duplication in tensor split layer assignment * generalize LLAMA_SPLIT_LAYER for all backends, do not expose device count and memory in llama.h * Only do device info print in the beginning and initialize one backend for cpu assist Add missing cleanup code * Rework backend memory management to make sure devices and buffers get properly allocated and freed * Rename cpu assist free function --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-07 06:54:50 +00:00
static size_t llama_get_device_count() {
cuda : rename build flag to LLAMA_CUDA (#6299)
2024-03-26 00:16:01 +00:00
#if defined(GGML_USE_CUDA)
Basic Vulkan Multi-GPU implementation (#5321) * Initial Vulkan multi-gpu implementation Move most global variables into backend context * Add names to backend device functions * Add further missing cleanup code * Reduce code duplication in tensor split layer assignment * generalize LLAMA_SPLIT_LAYER for all backends, do not expose device count and memory in llama.h * Only do device info print in the beginning and initialize one backend for cpu assist Add missing cleanup code * Rework backend memory management to make sure devices and buffers get properly allocated and freed * Rename cpu assist free function --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-07 06:54:50 +00:00
return ggml_backend_cuda_get_device_count();
Support multiple GPUs (split mode) on SYCL backend (#5806) * suport multiple cards: split-mode - layer|row * rm warning * rebase with master, support tow new OPs, close feature for -sm=row, fix for unit test * update news * fix merge error * update according to review comments
2024-03-02 11:49:30 +00:00
#elif defined(GGML_USE_SYCL)
return ggml_backend_sycl_get_device_count();
Basic Vulkan Multi-GPU implementation (#5321) * Initial Vulkan multi-gpu implementation Move most global variables into backend context * Add names to backend device functions * Add further missing cleanup code * Reduce code duplication in tensor split layer assignment * generalize LLAMA_SPLIT_LAYER for all backends, do not expose device count and memory in llama.h * Only do device info print in the beginning and initialize one backend for cpu assist Add missing cleanup code * Rework backend memory management to make sure devices and buffers get properly allocated and freed * Rename cpu assist free function --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-07 06:54:50 +00:00
#elif defined(GGML_USE_VULKAN)
return ggml_backend_vk_get_device_count();
#else
return 1;
#endif
}
static size_t llama_get_device_memory(int device) {
cuda : rename build flag to LLAMA_CUDA (#6299)
2024-03-26 00:16:01 +00:00
#if defined(GGML_USE_CUDA)
Basic Vulkan Multi-GPU implementation (#5321) * Initial Vulkan multi-gpu implementation Move most global variables into backend context * Add names to backend device functions * Add further missing cleanup code * Reduce code duplication in tensor split layer assignment * generalize LLAMA_SPLIT_LAYER for all backends, do not expose device count and memory in llama.h * Only do device info print in the beginning and initialize one backend for cpu assist Add missing cleanup code * Rework backend memory management to make sure devices and buffers get properly allocated and freed * Rename cpu assist free function --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-07 06:54:50 +00:00
size_t total;
size_t free;
Correct free memory and total memory. (#6630) Co-authored-by: MasterYi <zouxiaoyi@kylinos.cn>
2024-04-12 08:28:12 +00:00
ggml_backend_cuda_get_device_memory(device, &free, &total);
Basic Vulkan Multi-GPU implementation (#5321) * Initial Vulkan multi-gpu implementation Move most global variables into backend context * Add names to backend device functions * Add further missing cleanup code * Reduce code duplication in tensor split layer assignment * generalize LLAMA_SPLIT_LAYER for all backends, do not expose device count and memory in llama.h * Only do device info print in the beginning and initialize one backend for cpu assist Add missing cleanup code * Rework backend memory management to make sure devices and buffers get properly allocated and freed * Rename cpu assist free function --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-07 06:54:50 +00:00
return free;
Support multiple GPUs (split mode) on SYCL backend (#5806) * suport multiple cards: split-mode - layer|row * rm warning * rebase with master, support tow new OPs, close feature for -sm=row, fix for unit test * update news * fix merge error * update according to review comments
2024-03-02 11:49:30 +00:00
#elif defined(GGML_USE_SYCL)
size_t total;
size_t free;
Correct free memory and total memory. (#6630) Co-authored-by: MasterYi <zouxiaoyi@kylinos.cn>
2024-04-12 08:28:12 +00:00
ggml_backend_sycl_get_device_memory(device, &free, &total);
Support multiple GPUs (split mode) on SYCL backend (#5806) * suport multiple cards: split-mode - layer|row * rm warning * rebase with master, support tow new OPs, close feature for -sm=row, fix for unit test * update news * fix merge error * update according to review comments
2024-03-02 11:49:30 +00:00
return free;
Basic Vulkan Multi-GPU implementation (#5321) * Initial Vulkan multi-gpu implementation Move most global variables into backend context * Add names to backend device functions * Add further missing cleanup code * Reduce code duplication in tensor split layer assignment * generalize LLAMA_SPLIT_LAYER for all backends, do not expose device count and memory in llama.h * Only do device info print in the beginning and initialize one backend for cpu assist Add missing cleanup code * Rework backend memory management to make sure devices and buffers get properly allocated and freed * Rename cpu assist free function --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-07 06:54:50 +00:00
#elif defined(GGML_USE_VULKAN)
size_t total;
size_t free;
Correct free memory and total memory. (#6630) Co-authored-by: MasterYi <zouxiaoyi@kylinos.cn>
2024-04-12 08:28:12 +00:00
ggml_backend_vk_get_device_memory(device, &free, &total);
Basic Vulkan Multi-GPU implementation (#5321) * Initial Vulkan multi-gpu implementation Move most global variables into backend context * Add names to backend device functions * Add further missing cleanup code * Reduce code duplication in tensor split layer assignment * generalize LLAMA_SPLIT_LAYER for all backends, do not expose device count and memory in llama.h * Only do device info print in the beginning and initialize one backend for cpu assist Add missing cleanup code * Rework backend memory management to make sure devices and buffers get properly allocated and freed * Rename cpu assist free function --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-07 06:54:50 +00:00
return free;
#else
return 1;
GGML_UNUSED(device);
#endif
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
//
// globals
//
struct llama_state {
llama : set metal log callback correctly (#4204)
2023-11-24 17:10:01 +00:00
llama_state() {
#ifdef GGML_USE_METAL
metal : remove old API (#4919) ggml-ci
2024-01-13 18:45:45 +00:00
ggml_backend_metal_log_set_callback(log_callback, log_callback_user_data);
llama : set metal log callback correctly (#4204)
2023-11-24 17:10:01 +00:00
#endif
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
// We save the log callback globally
metal : reusing llama.cpp logging (#3152) * metal : reusing llama.cpp logging * cmake : build fix * metal : logging callback * metal : logging va_args memory fix * metal : minor cleanup * metal : setting function like logging macro to capital letters * llama.cpp : trailing whitespace fix * ggml : log level enum used by llama * Makefile : cleanup ggml-metal recipe * ggml : ggml_log_callback typedef * ggml : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-27 15:48:33 +00:00
ggml_log_callback log_callback = llama_log_callback_default;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
void * log_callback_user_data = nullptr;
};
static llama_state g_state;
// available llama models
enum e_model {
MODEL_UNKNOWN,
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
MODEL_17M,
MODEL_22M,
MODEL_33M,
MODEL_109M,
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
MODEL_137M,
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
MODEL_335M,
llama : add more qwen2 models (#5071)
2024-01-22 07:33:19 +00:00
MODEL_0_5B,
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
MODEL_1B,
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
MODEL_2B,
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
MODEL_3B,
llama : add more qwen2 models (#5071)
2024-01-22 07:33:19 +00:00
MODEL_4B,
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
MODEL_7B,
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
MODEL_8B,
llama : add StableLM2 12B (#6635) * StableLM2 12B support for huggingface -> GGUF * StableLM12 tensormapping and constants * StableLM-2-12b model support * fix * Added 12B support * Removed autoformatting; resolved bug where model_arch was not selecting StableLM2 * Formatting * Do QK norm stacking in model conversion step * Converge StableLM and StableLM2 code to simplify graph construction * Fix accidental removal * Removed warnings * Revert formatter * Move QK norm stack to private function so it's easier to read * refactor stablelm graph builder to support 1.6, 3b and 12b more efficiently * Proper check for None type for new_name to avoid crash; formatting; revert change to base class `write_tensors()` * Format * Formatting * format Co-authored-by: compilade <git@compilade.net> * Fix incorrect check for K norm * space after commas; Keep indentation multiple of 4 spaces * Flake8 format * Removed unnecessary conditional branches * Removed unused comment * Fixed incorrect tensor passing * Format --------- Co-authored-by: compilade <git@compilade.net>
2024-04-16 15:48:35 +00:00
MODEL_12B,
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
MODEL_13B,
llama : add support for Orion-14B (#5118) * add support for Orion-14B(https://huggingface.co/OrionStarAI/Orion-14B-Chat) * flake8 support * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * Update llama.cpp * Update llama.cpp --------- Co-authored-by: lixiaopu <lixiaopu@cmcm.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-01-28 08:00:30 +00:00
MODEL_14B,
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
MODEL_15B,
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
MODEL_20B,
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
MODEL_30B,
fix convert.py for codellama, add llama 34B to the list of recognized models (#2768)
2023-08-24 15:44:11 +00:00
MODEL_34B,
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
MODEL_35B,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
MODEL_40B,
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
MODEL_65B,
MODEL_70B,
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
MODEL_314B,
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
MODEL_SMALL,
MODEL_MEDIUM,
MODEL_LARGE,
MODEL_XL,
llama : add qwen2moe (#6074) * support qwen2moe * fix-review * metal : support unary ops for nelements % 4 != 0 * metal : require contiguousness for float4 unary kernels * metal : require contiguousness for float4 unary kernels (cont) * fix-review * names : for brevity "SHARED_EXP" -> "SHEXP" * llama : reuse build_moe_ffn() * llama : add model type name --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-16 15:40:48 +00:00
MODEL_A2_7B,
llama : add model types for mixtral (#6589)
2024-04-10 15:24:14 +00:00
MODEL_8x7B,
MODEL_8x22B,
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
MODEL_16x12B,
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
};
llama : fix data units (#4101) * llama : fix data units ggml-ci * Revert "llama : fix data units" This reverts commit f5feac831fe225ed7f3db938d115732a49dccfc4. * llama : disambiguate data units ggml-ci
2023-11-17 08:00:15 +00:00
static const size_t kiB = 1024;
static const size_t MiB = 1024*kiB;
static const size_t GiB = 1024*MiB;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
struct llama_hparams {
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
bool vocab_only;
bool rope_finetuned;
llama : allow gguf RoPE keys to be overridden with defaults (#3240)
2023-09-20 16:12:47 +00:00
uint32_t n_vocab;
uint32_t n_ctx_train; // context size the model was trained on
uint32_t n_embd;
uint32_t n_head;
uint32_t n_head_kv;
uint32_t n_layer;
uint32_t n_rot;
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
uint32_t n_embd_head_k; // dimension of keys (d_k). d_q is assumed to be the same, but there are n_head q heads, and only n_head_kv k-v heads
uint32_t n_embd_head_v; // dimension of values (d_v) aka n_embd_head
llama : allow gguf RoPE keys to be overridden with defaults (#3240)
2023-09-20 16:12:47 +00:00
uint32_t n_ff;
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
uint32_t n_expert = 0;
uint32_t n_expert_used = 0;
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
uint32_t n_vocab_type = 0; // for BERT-style token types
llama : allow gguf RoPE keys to be overridden with defaults (#3240)
2023-09-20 16:12:47 +00:00
float f_norm_eps;
float f_norm_rms_eps;
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
float rope_freq_base_train;
float rope_freq_scale_train;
uint32_t n_yarn_orig_ctx;
llama : add custom RoPE (#2054) * Implement customizable RoPE The original RoPE has pre-defined parameters theta_i = 10000^(−2(i−1)/d), for i in [1, 2, ..., d/2] Our customizable RoPE, ggml_rope_custom_inplace, uses theta_i = scale * base^(−2(i−1)/d), for i in [1, 2, ..., d/2] with the default matches the original scale = 1.0 base = 10000 The new command line arguments --rope-freq-base --rope-freq-scale set the two new RoPE parameter. Recent researches show changing these two parameters extends the context limit with minimal loss. 1. Extending Context to 8K kaiokendev https://kaiokendev.github.io/til#extending-context-to-8k 2. Extending Context Window of Large Language Models via Positional Interpolation Shouyuan Chen, Sherman Wong, Liangjian Chen, Yuandong Tian https://arxiv.org/abs/2306.15595 3. NTK-Aware Scaled RoPE allows LLaMA models to have extended (8k+) context size without any fine-tuning and minimal perplexity degradation. https://www.reddit.com/user/bloc97 https://www.reddit.com/r/LocalLLaMA/comments/14lz7j5/ntkaware_scaled_rope_allows_llama_models_to_have/ For the bold, try adding the following command line parameters to your favorite model: -c 16384 --rope-freq-base 80000 --rope-freq-scale 0.5 * ggml-metal: fix custom rope * common: fix argument names in help * llama: increase MEM_REQ_EVAL for MODEL_3B It avoids crashing for quantized weights on CPU. Better ways to calculate the required buffer size would be better. * llama: make MEM_REQ_EVAL depend on n_ctx * server: use proper Content-Type in curl examples Without the header Content-Type: application/json, curl will POST with Content-Type: application/x-www-form-urlencoded Though our simple server doesn't care, the httplib.h used has a limit with CPPHTTPLIB_FORM_URL_ENCODED_PAYLOAD_MAX_LENGTH 8192 With Content-Type: application/json, we can send large json data. * style : minor fixes, mostly indentations * ggml : fix asserts --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-15 10:34:16 +00:00
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
// for State Space Models
uint32_t ssm_d_conv = 0;
uint32_t ssm_d_inner = 0;
uint32_t ssm_d_state = 0;
uint32_t ssm_dt_rank = 0;
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
float f_clamp_kqv = 0.0f;
float f_max_alibi_bias = 0.0f;
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
float f_logit_scale = 0.0f;
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
bool causal_attn = true;
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
bool use_alibi = false; // currently, we need KQ_pos data for ALiBi-based models
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
llama : allow for user specified embedding pooling type (#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-03 10:40:27 +00:00
enum llama_pooling_type pooling_type = LLAMA_POOLING_TYPE_NONE;
enum llama_rope_type rope_type = LLAMA_ROPE_TYPE_NONE;
enum llama_rope_scaling_type rope_scaling_type_train = LLAMA_ROPE_SCALING_TYPE_NONE;
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
bool operator!=(const llama_hparams & other) const {
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
if (this->vocab_only != other.vocab_only) return true;
if (this->n_vocab != other.n_vocab) return true;
if (this->n_ctx_train != other.n_ctx_train) return true;
if (this->n_embd != other.n_embd) return true;
if (this->n_head != other.n_head) return true;
if (this->n_head_kv != other.n_head_kv) return true;
if (this->n_layer != other.n_layer) return true;
if (this->n_rot != other.n_rot) return true;
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
if (this->n_embd_head_k != other.n_embd_head_k) return true;
if (this->n_embd_head_v != other.n_embd_head_v) return true;
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
if (this->n_ff != other.n_ff) return true;
if (this->n_expert != other.n_expert) return true;
if (this->n_expert_used != other.n_expert_used) return true;
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
if (this->rope_finetuned != other.rope_finetuned) return true;
if (this->n_yarn_orig_ctx != other.n_yarn_orig_ctx) return true;
llama : correct hparams comparison (#3446) * fixed floating point comparison issues * updated implementation for hparam comparison to handle inf and NaN * fixed code review comments * minor simplification * rename is_float_eq -> is_float_close --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com>
2023-10-06 10:47:59 +00:00
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
if (this->ssm_d_conv != other.ssm_d_conv) return true;
if (this->ssm_d_inner != other.ssm_d_inner) return true;
if (this->ssm_d_state != other.ssm_d_state) return true;
if (this->ssm_dt_rank != other.ssm_dt_rank) return true;
cuda : improve cuda pool efficiency using virtual memory (#4606) * cuda : improve cuda pool efficiency using virtual memory * fix mixtral * fix cmake build * check for vmm support, disable for hip ggml-ci * fix hip build * clarify granularity * move all caps to g_device_caps * refactor error checking * add cuda_pool_alloc, refactor most pool allocations ggml-ci * fix hip build * CUBLAS_TF32_TENSOR_OP_MATH is not a macro * more hip crap * llama : fix msvc warnings * ggml : fix msvc warnings * minor * minor * cuda : fallback to CPU on host buffer alloc fail * Update ggml-cuda.cu Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * Update ggml-cuda.cu Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * ensure allocations are always aligned * act_size -> actual_size --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2023-12-24 13:34:22 +00:00
const float EPSILON = 1e-9f;
llama : correct hparams comparison (#3446) * fixed floating point comparison issues * updated implementation for hparam comparison to handle inf and NaN * fixed code review comments * minor simplification * rename is_float_eq -> is_float_close --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com>
2023-10-06 10:47:59 +00:00
sampling : refactor init to use llama_sampling_params (#3696) * sampling : refactor init to use llama_sampling_params * llama : combine repetition, frequency and presence penalties in 1 call * examples : remove embd-input and gptneox-wip * sampling : rename penalty params + reduce size of "prev" vector * sampling : add llama_sampling_print helper * sampling : hide prev behind API and apply #3661 ggml-ci
2023-10-20 18:07:23 +00:00
if (!is_float_close(this->f_norm_eps, other.f_norm_eps, EPSILON)) return true;
if (!is_float_close(this->f_norm_rms_eps, other.f_norm_rms_eps, EPSILON)) return true;
if (!is_float_close(this->rope_freq_base_train, other.rope_freq_base_train, EPSILON)) return true;
llama : correct hparams comparison (#3446) * fixed floating point comparison issues * updated implementation for hparam comparison to handle inf and NaN * fixed code review comments * minor simplification * rename is_float_eq -> is_float_close --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com>
2023-10-06 10:47:59 +00:00
if (!is_float_close(this->rope_freq_scale_train, other.rope_freq_scale_train, EPSILON)) return true;
return false;
llama : grouped-query attention + LLaMAv2 70B support (#2276) * CUDA: GQA implementation * llama : support for GQA and LLaMAv2 70B ggml-ci * py : fix hparams parsing (if-else blocks) ggml-ci * py : oh boy .. ggml-ci * help : fix gqa value for 70B ggml-ci --------- Co-authored-by: JohannesGaessler <johannesg@5d6.de>
2023-07-23 12:09:47 +00:00
}
uint32_t n_gqa() const {
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
if (n_head_kv == 0) {
return 0;
}
llama : grouped-query attention + LLaMAv2 70B support (#2276) * CUDA: GQA implementation * llama : support for GQA and LLaMAv2 70B ggml-ci * py : fix hparams parsing (if-else blocks) ggml-ci * py : oh boy .. ggml-ci * help : fix gqa value for 70B ggml-ci --------- Co-authored-by: JohannesGaessler <johannesg@5d6.de>
2023-07-23 12:09:47 +00:00
return n_head/n_head_kv;
}
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
uint32_t n_embd_k_gqa() const { // dimension of key embeddings across all k-v heads
return n_embd_head_k * n_head_kv;
llama : grouped-query attention + LLaMAv2 70B support (#2276) * CUDA: GQA implementation * llama : support for GQA and LLaMAv2 70B ggml-ci * py : fix hparams parsing (if-else blocks) ggml-ci * py : oh boy .. ggml-ci * help : fix gqa value for 70B ggml-ci --------- Co-authored-by: JohannesGaessler <johannesg@5d6.de>
2023-07-23 12:09:47 +00:00
}
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
uint32_t n_embd_v_gqa() const { // dimension of value embeddings across all k-v heads
return n_embd_head_v * n_head_kv;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
uint32_t n_embd_k_s() const { // dimension of the rolling state embeddings
// corresponds to Mamba's conv_states size
// TODO: maybe support other convolution strides than 1
// NOTE: since the first column of the conv_state is shifted out each time, it's not actually needed
return (ssm_d_conv > 0 ? ssm_d_conv - 1 : 0) * ssm_d_inner;
}
uint32_t n_embd_v_s() const { // dimension of the recurrent state embeddings
// corresponds to Mamba's ssm_states size
return ssm_d_state * ssm_d_inner;
}
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
};
llama : optimize memory buffers (#2325)
2023-07-22 18:17:57 +00:00
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
struct llama_cparams {
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
uint32_t n_ctx; // context size used during inference
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
uint32_t n_batch;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
uint32_t n_ubatch;
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
uint32_t n_seq_max;
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
uint32_t n_threads; // number of threads to use for generation
uint32_t n_threads_batch; // number of threads to use for batch processing
llama : allow for user specified embedding pooling type (#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-03 10:40:27 +00:00
float rope_freq_base;
float rope_freq_scale;
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
uint32_t n_yarn_orig_ctx;
// These hyperparameters are not exposed in GGUF, because all
// existing YaRN models use the same values for them.
float yarn_ext_factor;
float yarn_attn_factor;
float yarn_beta_fast;
float yarn_beta_slow;
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
float defrag_thold;
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
bool embeddings;
llama : add support for GritLM (#5959) * add gritlm example * gritlm results match * tabs to spaces * comment out debug printing * rebase to new embed * gritlm embeddings are back babeee * add to gitignore * allow to toggle embedding mode * Clean-up GritLM sample code. * Fix types. * Flush stdout and output ending newline if streaming. * mostly style fixes; correct KQ_mask comment * add causal_attn flag to llama_cparams * gritml : minor * llama : minor --------- Co-authored-by: Douglas Hanley <thesecretaryofwar@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-10 15:56:30 +00:00
bool causal_attn;
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
bool offload_kqv;
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
bool flash_attn;
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
llama : allow for user specified embedding pooling type (#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-03 10:40:27 +00:00
enum llama_pooling_type pooling_type;
backend : add eval callback (#4935) * backend : add eval callback ggml-ci * backend : group nodes in a single compute when user don't need them * backend : clean-up the implementation ggml-ci * simple : do not perform tensor data copy if not needed * simple : fix * simple : no need for ggml_is_contiguous + fix bool parse * llama : fix callback placement in llama_context_params * backend : avoid double-ask callback calls * simple : restore examples, imatrix will serve as a demo
2024-01-17 16:39:41 +00:00
ggml_backend_sched_eval_callback cb_eval;
void * cb_eval_user_data;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
};
struct llama_layer {
// normalization
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
struct ggml_tensor * attn_norm;
struct ggml_tensor * attn_norm_b;
struct ggml_tensor * attn_norm_2;
struct ggml_tensor * attn_norm_2_b;
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
struct ggml_tensor * attn_q_norm;
struct ggml_tensor * attn_q_norm_b;
struct ggml_tensor * attn_k_norm;
struct ggml_tensor * attn_k_norm_b;
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
struct ggml_tensor * attn_out_norm;
struct ggml_tensor * attn_out_norm_b;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
// attention
struct ggml_tensor * wq;
struct ggml_tensor * wk;
struct ggml_tensor * wv;
struct ggml_tensor * wo;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
struct ggml_tensor * wqkv;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
// attention bias
llama : support attention bias on LLaMA architecture (#4283) * Support attention_bias on LLaMA architecture QKVO bias, should fix InternLM (https://github.com/ggerganov/llama.cpp/issues/3133) and works for LLaMAfied Qwen models (https://github.com/ggerganov/llama.cpp/pull/3743#issuecomment-1825923608). * check existence of qkvo bias while loading llama models Tested on LLaMA2, CUDA and CPU. * Update llama.cpp
2023-12-01 18:17:06 +00:00
struct ggml_tensor * bq;
struct ggml_tensor * bk;
struct ggml_tensor * bv;
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
struct ggml_tensor * bo;
struct ggml_tensor * bqkv;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
// normalization
struct ggml_tensor * ffn_norm;
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
struct ggml_tensor * ffn_norm_b;
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
struct ggml_tensor * layer_out_norm;
struct ggml_tensor * layer_out_norm_b;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
// ff
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
struct ggml_tensor * ffn_gate; // w1
struct ggml_tensor * ffn_down; // w2
struct ggml_tensor * ffn_up; // w3
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
// ff MoE
struct ggml_tensor * ffn_gate_inp;
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
struct ggml_tensor * ffn_gate_exps;
struct ggml_tensor * ffn_down_exps;
struct ggml_tensor * ffn_up_exps ;
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
llama : add qwen2moe (#6074) * support qwen2moe * fix-review * metal : support unary ops for nelements % 4 != 0 * metal : require contiguousness for float4 unary kernels * metal : require contiguousness for float4 unary kernels (cont) * fix-review * names : for brevity "SHARED_EXP" -> "SHEXP" * llama : reuse build_moe_ffn() * llama : add model type name --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-16 15:40:48 +00:00
// ff shared expert (shexp)
struct ggml_tensor * ffn_gate_inp_shexp;
struct ggml_tensor * ffn_gate_shexp;
struct ggml_tensor * ffn_down_shexp;
struct ggml_tensor * ffn_up_shexp;
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
// ff bias
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
struct ggml_tensor * ffn_down_b; // b2
struct ggml_tensor * ffn_up_b; // b3
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
struct ggml_tensor * ffn_act;
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
// mamba proj
struct ggml_tensor * ssm_in;
struct ggml_tensor * ssm_x;
struct ggml_tensor * ssm_dt;
struct ggml_tensor * ssm_out;
// mamba
struct ggml_tensor * ssm_conv1d;
struct ggml_tensor * ssm_a;
struct ggml_tensor * ssm_d;
// mamba bias
struct ggml_tensor * ssm_conv1d_b;
struct ggml_tensor * ssm_dt_b;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
};
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
struct llama_kv_cell {
llama_pos pos = -1;
llama_pos delta = 0;
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
int32_t src = 0; // used by recurrent state models to copy states
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
std::set<llama_seq_id> seq_id;
bool has_seq_id(const llama_seq_id & id) const {
return seq_id.find(id) != seq_id.end();
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
bool is_empty() const {
return seq_id.empty();
}
bool is_same_seq(const llama_kv_cell & other) const {
return seq_id == other.seq_id;
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
};
// ring-buffer of cached KV data
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
struct llama_kv_cache {
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
bool has_shift = false;
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
bool do_defrag = false;
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
bool do_copy = false;
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
bool recurrent = false; // with recurrent state models, a cell can hold the state for more than one past token
bool v_trans = true; // the value tensor is transposed
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
kv cache slot search improvements (#3493) * kv cache slot search improvements * Use n_ctx in kv find slot for consistency * Ensure kv cache head points to a valid slot in llama_decode internal * Add some comments to prevent dumb people (like me) from getting confused.
2023-10-06 16:10:13 +00:00
// Note: The value of head isn't only used to optimize searching
// for a free KV slot. llama_decode_internal also uses it, so it
// cannot be freely changed after a slot has been allocated.
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
uint32_t head = 0;
uint32_t size = 0;
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
uint32_t used = 0; // used cells (i.e. at least one seq_id)
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
// computed before each graph build
uint32_t n = 0;
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ggml_type type_k = GGML_TYPE_F16;
ggml_type type_v = GGML_TYPE_F16;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
std::vector<llama_kv_cell> cells;
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
std::vector<struct ggml_tensor *> k_l; // per layer
std::vector<struct ggml_tensor *> v_l;
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
std::vector<struct ggml_context *> ctxs;
std::vector<ggml_backend_buffer_t> bufs;
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
size_t total_size() const {
size_t size = 0;
for (ggml_backend_buffer_t buf : bufs) {
size += ggml_backend_buffer_get_size(buf);
}
return size;
}
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
~llama_kv_cache() {
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (struct ggml_context * ctx : ctxs) {
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
ggml_free(ctx);
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (ggml_backend_buffer_t buf : bufs) {
ggml_backend_buffer_free(buf);
}
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
};
llama : add support for control vectors (#5970) * control vector api and implementation * control-vectors : minor code style updates * disable control vector when data == nullptr use -1 for disabled range (also on init) in case we ever support controlling layer 0 (embeddings) --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-15 20:43:02 +00:00
struct llama_control_vector {
std::vector<struct ggml_tensor *> tensors; // per layer
std::vector<struct ggml_context *> ctxs;
std::vector<ggml_backend_buffer_t> bufs;
int32_t layer_start = -1;
int32_t layer_end = -1;
ggml_tensor * tensor_for(int il) const {
if (il < 0 || il < layer_start || il > layer_end || (size_t) il >= tensors.size()) {
return nullptr;
}
return tensors[il];
}
~llama_control_vector() {
for (struct ggml_context * ctx : ctxs) {
ggml_free(ctx);
}
for (ggml_backend_buffer_t buf : bufs) {
ggml_backend_buffer_free(buf);
}
}
};
llama : make model stateless and context stateful (llama_state) (#1797) * llama : make model stateless and context stateful * llama : minor cleanup * llama : update internal API declaration * Apply suggestions from code review fix style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Missing model memory release * Fix style * Add deprecated warning for public API function llama_init_from_file * Update public API use cases: move away from deprecated llama_init_from_file * Deprecate public API function llama_apply_lora_from_file --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-24 08:47:58 +00:00
struct llama_vocab {
using id = int32_t;
using token = std::string;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
using ttype = llama_token_type;
llama : make model stateless and context stateful (llama_state) (#1797) * llama : make model stateless and context stateful * llama : minor cleanup * llama : update internal API declaration * Apply suggestions from code review fix style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Missing model memory release * Fix style * Add deprecated warning for public API function llama_init_from_file * Update public API use cases: move away from deprecated llama_init_from_file * Deprecate public API function llama_apply_lora_from_file --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-24 08:47:58 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
struct token_data {
token text;
llama : make model stateless and context stateful (llama_state) (#1797) * llama : make model stateless and context stateful * llama : minor cleanup * llama : update internal API declaration * Apply suggestions from code review fix style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Missing model memory release * Fix style * Add deprecated warning for public API function llama_init_from_file * Update public API use cases: move away from deprecated llama_init_from_file * Deprecate public API function llama_apply_lora_from_file --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-24 08:47:58 +00:00
float score;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
ttype type;
llama : make model stateless and context stateful (llama_state) (#1797) * llama : make model stateless and context stateful * llama : minor cleanup * llama : update internal API declaration * Apply suggestions from code review fix style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Missing model memory release * Fix style * Add deprecated warning for public API function llama_init_from_file * Update public API use cases: move away from deprecated llama_init_from_file * Deprecate public API function llama_apply_lora_from_file --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-24 08:47:58 +00:00
};
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
enum llama_vocab_type type = LLAMA_VOCAB_TYPE_SPM;
enum llama_vocab_pre_type type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : make model stateless and context stateful (llama_state) (#1797) * llama : make model stateless and context stateful * llama : minor cleanup * llama : update internal API declaration * Apply suggestions from code review fix style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Missing model memory release * Fix style * Add deprecated warning for public API function llama_init_from_file * Update public API use cases: move away from deprecated llama_init_from_file * Deprecate public API function llama_apply_lora_from_file --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-24 08:47:58 +00:00
std::unordered_map<token, id> token_to_id;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
std::vector<token_data> id_to_token;
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
std::unordered_map<token, id> special_tokens_cache;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
std::map<std::pair<std::string, std::string>, int> bpe_ranks;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
// default LLaMA special tokens
BERT tokenizer fixes (#6498) Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
2024-04-09 17:44:08 +00:00
id special_bos_id = 1;
id special_eos_id = 2;
id special_unk_id = 0;
id special_sep_id = -1;
id special_pad_id = -1;
id special_cls_id = -1;
id special_mask_id = -1;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
Respect tokenizer.ggml.add_bos_token value when tokenizing (#4040) * gguf-py: gguf-dump: Respect --no-tensor flag in JSON mode. * Respect add_bos_token GGUF metadata value * gguf-py: Try to fix SpecialVocab giving up too easily for the Nth time
2023-11-17 02:14:37 +00:00
int special_add_bos = -1; // -1 unknown, 1 add, 0 don't add.
int special_add_eos = -1; // -1 unknown, 1 add, 0 don't add.
sampling : refactor init to use llama_sampling_params (#3696) * sampling : refactor init to use llama_sampling_params * llama : combine repetition, frequency and presence penalties in 1 call * examples : remove embd-input and gptneox-wip * sampling : rename penalty params + reduce size of "prev" vector * sampling : add llama_sampling_print helper * sampling : hide prev behind API and apply #3661 ggml-ci
2023-10-20 18:07:23 +00:00
id linefeed_id = 13;
gguf : add special tokens metadata for FIM/Infill (#6689) This commit adds special token metadata for Fill-In-the-Middle (FIM)/Infill to the GGUF model. The motivation for this is that currently there is support for CodeLlama but other models exist now like CodeGemma, but the different models use different token ids for the special tokens and this commit allows for supporting multiple models. Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
2024-04-16 06:13:13 +00:00
id special_prefix_id = -1;
id special_suffix_id = -1;
id special_middle_id = -1;
llama : support Llama 3 HF conversion (#6745) * Support Llama 3 conversion The tokenizer is BPE. * style * Accept suggestion Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> * llama : add llama_token_is_eog() ggml-ci * llama : auto-detect more EOT tokens when missing in KV data * convert : replacing EOS token is a hack * llama : fix codegemma EOT token + add TODOs * llama : fix model type string for 8B model --------- Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-21 11:50:41 +00:00
id special_eot_id = -1; // TODO: move above after "eos_id", and here add "file separator" token
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
bool add_space_prefix = true;
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
int find_bpe_rank(const std::string & token_left, const std::string & token_right) const {
GGML_ASSERT(token_left.find(' ') == std::string::npos);
GGML_ASSERT(token_left.find('\n') == std::string::npos);
GGML_ASSERT(token_right.find(' ') == std::string::npos);
GGML_ASSERT(token_right.find('\n') == std::string::npos);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
auto it = bpe_ranks.find(std::make_pair(token_left, token_right));
if (it == bpe_ranks.end()) {
return -1;
}
return it->second;
}
llama : make model stateless and context stateful (llama_state) (#1797) * llama : make model stateless and context stateful * llama : minor cleanup * llama : update internal API declaration * Apply suggestions from code review fix style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Missing model memory release * Fix style * Add deprecated warning for public API function llama_init_from_file * Update public API use cases: move away from deprecated llama_init_from_file * Deprecate public API function llama_apply_lora_from_file --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-24 08:47:58 +00:00
};
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
struct llama_model {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
e_model type = MODEL_UNKNOWN;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llm_arch arch = LLM_ARCH_UNKNOWN;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama_ftype ftype = LLAMA_FTYPE_ALL_F32;
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
std::string name = "n/a";
llama : allow gguf RoPE keys to be overridden with defaults (#3240)
2023-09-20 16:12:47 +00:00
llama_hparams hparams = {};
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama_vocab vocab;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
struct ggml_tensor * tok_embd;
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
struct ggml_tensor * type_embd;
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
struct ggml_tensor * pos_embd;
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
struct ggml_tensor * tok_norm;
struct ggml_tensor * tok_norm_b;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
struct ggml_tensor * output_norm;
struct ggml_tensor * output_norm_b;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
struct ggml_tensor * output;
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
struct ggml_tensor * output_b;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
std::vector<llama_layer> layers;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
llama_split_mode split_mode;
int main_gpu;
Multi GPU support, CUDA refactor, CUDA scratch buffer (#1703) * CUDA multi GPU + scratch ggml_cuda_compute_forward Tensor parallelism ggml_cuda_add ggml_cuda_rms_norm ggml_cuda_silu CUDA scratch buffer --main-gpu CLI option
2023-06-06 19:33:23 +00:00
int n_gpu_layers;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : add functions to get the model's metadata (#4013) * llama : add functions to get the model's metadata * format -> std::to_string * better documentation
2023-11-17 15:17:37 +00:00
// gguf metadata
std::unordered_map<std::string, std::string> gguf_kv;
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// layer -> buffer type mapping
struct layer_buft {
layer_buft() : buft_matrix(nullptr), buft(nullptr) {}
layer_buft(ggml_backend_buffer_type_t matrix) : buft_matrix(matrix), buft(matrix) {}
layer_buft(ggml_backend_buffer_type_t matrix, ggml_backend_buffer_type_t other) : buft_matrix(matrix), buft(other) {}
ggml_backend_buffer_type_t buft_matrix; // matrices only - used by split buffers and backends that support only matrix multiplication
ggml_backend_buffer_type_t buft; // everything else
};
layer_buft buft_input;
layer_buft buft_output;
std::vector<layer_buft> buft_layer;
// contexts where the model tensors metadata is stored
std::vector<struct ggml_context *> ctxs;
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// the model memory buffers for the tensor data
std::vector<ggml_backend_buffer_t> bufs;
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
// model memory mapped files
llama_mmaps mappings;
Unmap the file in llama_free
2023-03-29 06:31:26 +00:00
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
// objects representing data potentially being locked in memory
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
llama_mlocks mlock_bufs;
llama_mlocks mlock_mmaps;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
// for quantize-stats only
std::vector<std::pair<std::string, struct ggml_tensor *>> tensors_by_name;
llama : make model stateless and context stateful (llama_state) (#1797) * llama : make model stateless and context stateful * llama : minor cleanup * llama : update internal API declaration * Apply suggestions from code review fix style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Missing model memory release * Fix style * Add deprecated warning for public API function llama_init_from_file * Update public API use cases: move away from deprecated llama_init_from_file * Deprecate public API function llama_apply_lora_from_file --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-24 08:47:58 +00:00
int64_t t_load_us = 0;
int64_t t_start_us = 0;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
~llama_model() {
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (struct ggml_context * ctx : ctxs) {
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
ggml_free(ctx);
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (ggml_backend_buffer_t buf : bufs) {
cuda : rename build flag to LLAMA_CUDA (#6299)
2024-03-26 00:16:01 +00:00
#ifdef GGML_USE_CUDA
backend : offload large batches to GPU (#6083) * backend : offload large batches to GPU * fix hip * code cleanup * fix CUDA split buffers * Update ggml-backend-impl.h Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix memset without set_device * imatrix : remove sched affix from weight names * sched : add a new split if the current one has too many inputs reduce max inputs per split more cleanup * update backends ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-03-18 10:03:04 +00:00
if (ggml_backend_buffer_get_type(buf) == ggml_backend_cpu_buffer_type()) {
ggml_backend_cuda_unregister_host_buffer(ggml_backend_buffer_get_base(buf));
}
#endif
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ggml_backend_buffer_free(buf);
}
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
};
struct llama_context {
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
llama_context(const llama_model & model) : model(model), t_start_us(model.t_start_us), t_load_us(model.t_load_us) {}
metal : release buffers when freeing metal context (#2062)
2023-07-01 18:14:59 +00:00
~llama_context() {
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ggml_backend_sched_free(sched);
for (ggml_backend_t backend : backends) {
ggml_backend_free(backend);
}
llama : pre-allocate input tensors in a separate buffer (#5100)
2024-01-24 11:48:14 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
ggml_backend_buffer_free(buf_output);
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
}
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
llama_cparams cparams;
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
std::vector<ggml_backend_t> backends;
#ifdef GGML_USE_METAL
ggml_backend_t backend_metal = nullptr;
#endif
ggml_backend_t backend_cpu = nullptr;
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
const llama_model & model;
// key + value cache for the self attention
struct llama_kv_cache kv_self;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
std::mt19937 rng;
Make loading weights 10-100x faster This is a breaking change that's going to give you three benefits: 1. Your inference commands should load 100x faster 2. You may be able to safely load models 2x larger 3. You can run many concurrent inference processes This was accomplished by changing the file format so we can mmap() weights directly into memory without having to read() or copy them thereby ensuring the kernel can make its file cache pages directly accessible to our inference processes; and secondly, that the file cache pages are much less likely to get evicted (which would force loads to hit disk) because they're no longer competing with memory pages that were needlessly created by gigabytes of standard i/o. The new file format supports single-file models like LLaMA 7b, and it also supports multi-file models like LLaMA 13B. Our Python tool now merges the foo.1, foo.2, etc. files back into a single file so that the C++ code which maps it doesn't need to reshape data every time. That's made llama.cpp so much simpler. Much of its load code has now been deleted. Furthermore, this change ensures that tensors are aligned properly on a 32-byte boundary. That opens the door to seeing if we can get additional performance gains on some microprocessors, by using ops that require memory alignment. Lastly note that both POSIX and the Windows platform are supported Fixes #91
2023-03-29 20:51:37 +00:00
bool has_evaluated_once = false;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
int64_t t_start_us;
int64_t t_load_us;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
int64_t t_sample_us = 0;
Add timings for the prompt evaluation (#478)
2023-03-25 14:34:23 +00:00
int64_t t_p_eval_us = 0;
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
int64_t t_eval_us = 0;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
int64_t t_compute_start_us = 0;
int64_t n_queued_tokens = 0;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
int32_t n_sample = 0; // number of tokens sampled
Add timings for the prompt evaluation (#478)
2023-03-25 14:34:23 +00:00
int32_t n_p_eval = 0; // number of tokens in eval calls for the prompt (with batch size > 1)
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
int32_t n_eval = 0; // number of eval calls
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// host buffer for the model output (logits and embeddings)
ggml_backend_buffer_t buf_output = nullptr;
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
// decode output (2-dimensional array: [n_outputs][n_vocab])
size_t logits_size = 0; // capacity (of floats) for logits
float * logits = nullptr;
std::vector<int32_t> output_ids; // map batch token positions to ids of the logits and embd buffers
size_t output_size = 0; // capacity (of tokens positions) for the output buffers
llama : support negative ith in llama_get_ API (#6519) * llama_sampling_sample with default args is more naively usable * Batches populated by either llama_batch_get_one or llama_batch_add work with default args * Previously get_one could use the default argument * Previously add should usually have used the last index where logits[idx] == true * This hopefully encourages the use of llama_batch_add * By giving expected results when using default arguments. * Adds "negative indexing" feature to llama_get_logits_ith and llama_get_embeddings_ith * Believed to work with any currently well behaved program * Default arg now works for both cases (previously would give strange results for add case) * Any non-negative number is unaffected and behaves as previously * Negative arguments were previously invalid. * Implemented as a special case of indexing as suggested by @compilade in https://github.com/ggerganov/llama.cpp/pull/6519 * Fixed mismatch type errors * cited in macOS CI tests * Missed in original updates based on PR feedback in https://github.com/ggerganov/llama.cpp/pull/6519
2024-04-08 13:02:30 +00:00
int32_t n_outputs = 0; // number of actually-used outputs in the current ubatch or last logical batch
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
bool logits_all = false;
Add embedding mode with arg flag. Currently working (#282) * working but ugly * add arg flag, not working on embedding mode * typo * Working! Thanks to @nullhook * make params argument instead of hardcoded boolean. remove useless time check * start doing the instructions but not finished. This probably doesnt compile * Embeddings extraction support --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-03-24 15:05:13 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
// embeddings output (2-dimensional array: [n_outputs][n_embd])
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
// populated only when pooling_type == LLAMA_POOLING_TYPE_NONE
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
size_t embd_size = 0; // capacity (of floats) for embeddings
float * embd = nullptr;
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
// sequence embeddings output (map of [n_embd] vectors)
// populated only when pooling_type != LLAMA_POOLING_TYPE_NONE
std::map<llama_seq_id, std::vector<float>> embd_seq;
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
// memory buffers used to evaluate the model
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
std::vector<uint8_t> buf_compute_meta;
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ggml_backend_sched_t sched = nullptr;
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
llama : add abort_callback to interrupt computation (#5409) * using abort_callback from ggml to stop llama computation * format fix * a brief explaining comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-02 19:52:25 +00:00
ggml_abort_callback abort_callback = nullptr;
void * abort_callback_data = nullptr;
llama : pre-allocate input tensors in a separate buffer (#5100)
2024-01-24 11:48:14 +00:00
// input tensors
struct ggml_tensor * inp_tokens; // I32 [n_batch]
struct ggml_tensor * inp_embd; // F32 [n_embd, n_batch]
struct ggml_tensor * inp_pos; // I32 [n_batch]
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
struct ggml_tensor * inp_out_ids; // I32 [n_outputs]
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
struct ggml_tensor * inp_KQ_mask; // F32 [kv_size, n_batch]
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
struct ggml_tensor * inp_KQ_pos; // F32 [n_kv]
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
struct ggml_tensor * inp_K_shift; // I32 [kv_size]
Use correct type of pooling for embedding models (#5500) Use correct type of pooling for embedding models
2024-02-15 17:21:49 +00:00
struct ggml_tensor * inp_mean; // F32 [n_batch, n_batch]
struct ggml_tensor * inp_cls; // I32 [n_batch]
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
struct ggml_tensor * inp_s_copy; // I32 [kv_size]
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
struct ggml_tensor * inp_s_mask; // F32 [1, n_kv]
struct ggml_tensor * inp_s_seq; // I32 [n_kv, n_batch]
llama : Metal inference (#1642) * mtl : export the LLaMA computation graph * ci : disable temporary * mtl : adapt the MNIST example as starter * mtl : no need for mtl-export tool, add cli arg for main instead * mtl : export just a small part of the graph for now to make it easier * mtl : move MSL code into separate file for easy editing * mtl : initial get_rows_q4_0 kernel * mtl : confirmed get_rows_q4_0 is working correctly * mtl : add rms_norm kernel + confirm working * mtl : add mul kernel + confirm working * mtl : initial mul_mat Q4 kernel (wrong results) * mtl : mul_mat fixes (still wrong) * mtl : another mul_mat Q4 (still does not work) * mtl : working mul_mat q4 * ggml : fix handling of "view" ops in ggml_graph_import() * mtl : add rope kernel * mtl : add reshape and transpose handling * ggml : store offset as opt arg for ggml_view_xd() operators * mtl : add cpy kernel + handle view ops * mtl : confirm f16 x f32 attention mul mat * mtl : add scale kernel * mtl : add diag_mask_inf kernel * mtl : fix soft_max kernel * ggml : update ggml_nbytes() to handle non-contiguous tensors * mtl : verify V tensor contents * mtl : add f32 -> f32 cpy kernel * mtl : add silu kernel * mtl : add non-broadcast mul kernel * mtl : full GPU inference of the computation graph * mtl : optimize rms_norm and soft_max kernels * mtl : add f16 mat x f32 vec multiplication kernel * mtl : fix bug in f16 x f32 mul mat + speed-up computation * mtl : faster mul_mat_q4_0_f32 kernel * mtl : fix kernel signature + roll inner loop * mtl : more threads for rms_norm + better timing * mtl : remove printfs from inner loop * mtl : simplify implementation * mtl : add save/load vocab to ggml file * mtl : plug Metal inference into llama.cpp (very quick-n-dirty) * mtl : make it work with main example Lots of hacks but at least now it generates text * mtl : preparing for merge * mtl : clean-up ggml mtl interface + suport scratch / inplace * mtl : remove temp / debug code * metal : final refactoring and simplification * Revert "ci : disable temporary" This reverts commit 98c267fc77fe811082f672538fc91bcfc9072d63. * metal : add comments * metal : clean-up stuff, fix typos * readme : add Metal instructions * readme : add example for main
2023-06-04 20:34:30 +00:00
llama : add support for control vectors (#5970) * control vector api and implementation * control-vectors : minor code style updates * disable control vector when data == nullptr use -1 for disabled range (also on init) in case we ever support controlling layer 0 (embeddings) --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-15 20:43:02 +00:00
// control vectors
struct llama_control_vector cvec;
mpi : add support for distributed inference via MPI (#2099) * MPI support, first cut * fix warnings, update README * fixes * wrap includes * PR comments * Update CMakeLists.txt * Add GH workflow, fix test * Add info to README * mpi : trying to move more MPI stuff into ggml-mpi (WIP) (#2099) * mpi : add names for layer inputs + prep ggml_mpi_graph_compute() * mpi : move all MPI logic into ggml-mpi Not tested yet * mpi : various fixes - communication now works but results are wrong * mpi : fix output tensor after MPI compute (still not working) * mpi : fix inference * mpi : minor * Add OpenMPI to GH action * [mpi] continue-on-error: true * mpi : fix after master merge * [mpi] Link MPI C++ libraries to fix OpenMPI * tests : fix new llama_backend API * [mpi] use MPI_INT32_T * mpi : factor out recv / send in functions and reuse * mpi : extend API to allow usage with outer backends (e.g. Metal) --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-10 15:49:56 +00:00
#ifdef GGML_USE_MPI
ggml_mpi_context * ctx_mpi = NULL;
#endif
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
};
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
//
// kv cache helpers
//
static bool llama_kv_cache_init(
struct llama_kv_cache & cache,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
const llama_context * ctx,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ggml_type type_k,
ggml_type type_v,
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
uint32_t kv_size,
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
bool offload) {
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
const llama_model & model = ctx->model;
const llama_cparams & cparams = ctx->cparams;
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
const struct llama_hparams & hparams = model.hparams;
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa() + hparams.n_embd_k_s();
const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa() + hparams.n_embd_v_s();
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
const int64_t n_layer = hparams.n_layer;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
cache.has_shift = false;
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
// TODO: find a nicer way to add other recurrent model architectures
cache.recurrent = model.arch == LLM_ARCH_MAMBA;
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cache.v_trans = !cparams.flash_attn;
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
chore: fix typo in llama.cpp (#7032) Co-authored-by: Jared Van Bortel <jared@nomic.ai>
2024-05-02 15:56:41 +00:00
// TODO: support mixed recurrent Transformer architectures
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
// NOTE: (!a || b) is a logical implication (a -> b)
GGML_ASSERT(!cache.recurrent || n_embd_k_gqa == hparams.n_embd_k_s());
GGML_ASSERT(!cache.recurrent || n_embd_v_gqa == hparams.n_embd_v_s());
GGML_ASSERT( cache.recurrent || n_embd_k_gqa == hparams.n_embd_k_gqa());
GGML_ASSERT( cache.recurrent || n_embd_v_gqa == hparams.n_embd_v_gqa());
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
cache.head = 0;
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
cache.size = kv_size;
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
cache.used = 0;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
cache.type_k = type_k;
cache.type_v = type_v;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
cache.cells.clear();
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
cache.cells.resize(kv_size);
if (cache.recurrent) {
// init state copy sources
for (uint32_t i = 0; i < cache.size; ++i) {
cache.cells[i].src = i;
}
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
#ifdef GGML_USE_CLBLAST
offload = false;
#endif
add log_callback to llama_context_params for custom logging. (#2234) * add log_callback to llama_context_params for custom logging. * Fix macro expansion on gcc * Add struct llama_state for global variables and move log_callback there * Turn log level into enum and some minor changes. * Remove model_for_logging parameter (not needed anymore) * Convert remaining fprintf(stderr, ...) calls to use new macros. * Fix enum and initialize g_state * Fix log calls after merge * Fix missing static * Add back all the new lines in the logging strings * Add comment for llama_log_callback and replace remaining printf calls --------- Co-authored-by: grahameth <-> Co-authored-by: Helmut <helmut.buhler@inf.h-brs.de>
2023-08-09 20:46:40 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// count used buffer types
std::map<ggml_backend_buffer_type_t, int> buft_layer_count;
if (offload) {
for (int64_t i = 0; i < n_layer; ++i) {
buft_layer_count[model.buft_layer[i].buft]++;
}
} else {
buft_layer_count[llama_default_buffer_type_cpu(true)] = n_layer;
}
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// create a context for each buffer type
std::map<ggml_backend_buffer_type_t, ggml_context *> ctx_map;
for (auto & it : buft_layer_count) {
int n_layers = it.second;
struct ggml_init_params params = {
/*.mem_size =*/ 2u*n_layers*ggml_tensor_overhead(),
/*.mem_buffer =*/ NULL,
/*.no_alloc =*/ true,
};
ggml_context * ctx = ggml_init(params);
if (!ctx) {
LLAMA_LOG_ERROR("%s: failed to allocate context for kv cache\n", __func__);
return false;
}
ctx_map[it.first] = ctx;
cache.ctxs.push_back(ctx);
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
cache.k_l.reserve(n_layer);
cache.v_l.reserve(n_layer);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
for (int i = 0; i < (int) n_layer; i++) {
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
struct ggml_context * ctx = offload ? ctx_map.at(model.buft_layer[i].buft) : cache.ctxs.front();
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
ggml_tensor * k = ggml_new_tensor_1d(ctx, type_k, n_embd_k_gqa*kv_size);
ggml_tensor * v = ggml_new_tensor_1d(ctx, type_v, n_embd_v_gqa*kv_size);
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
ggml_format_name(k, "cache_k_l%d", i);
ggml_format_name(v, "cache_v_l%d", i);
cache.k_l.push_back(k);
cache.v_l.push_back(v);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// allocate tensors and initialize the buffers to avoid NaNs in the padding
for (auto it : ctx_map) {
ggml_backend_buffer_type_t buft = it.first;
ggml_context * ctx = it.second;
ggml_backend_buffer_t buf = ggml_backend_alloc_ctx_tensors_from_buft(ctx, buft);
if (!buf) {
LLAMA_LOG_ERROR("%s: failed to allocate buffer for kv cache\n", __func__);
return false;
}
ggml_backend_buffer_clear(buf, 0);
LLAMA_LOG_INFO("%s: %10s KV buffer size = %8.2f MiB\n", __func__, ggml_backend_buffer_name(buf), ggml_backend_buffer_get_size(buf)/1024.0/1024.0);
cache.bufs.push_back(buf);
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
return true;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
// find an empty slot of size "n_tokens" in the cache
// updates the cache head
kv cache slot search improvements (#3493) * kv cache slot search improvements * Use n_ctx in kv find slot for consistency * Ensure kv cache head points to a valid slot in llama_decode internal * Add some comments to prevent dumb people (like me) from getting confused.
2023-10-06 16:10:13 +00:00
// Note: On success, it's important that cache.head points
// to the first cell of the slot.
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
static bool llama_kv_cache_find_slot(
llama : fix session saving/loading (#3400) * llama : fix session saving/loading * llama : temp fix for clearing "future" tokens from the KV cache * llama : fix handling of "future" tokens when loading sessions * llama : fix comments for llama_kv_cache API
2023-10-03 18:04:01 +00:00
struct llama_kv_cache & cache,
const struct llama_batch & batch) {
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
const uint32_t n_ctx = cache.size;
const uint32_t n_tokens = batch.n_tokens;
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
if (cache.recurrent) {
// For recurrent state architectures (like Mamba),
// each KV cache cell can store the state for a whole sequence.
llama_seq_id min = cache.size - 1;
llama_seq_id max = 0;
for (uint32_t i = 0; i < n_tokens; ++i) {
for (int32_t j = 0; j < batch.n_seq_id[i]; ++j) {
llama_seq_id seq_id = batch.seq_id[i][j];
// make sure it's a valid seq_id
if ((uint32_t) seq_id < cache.size) {
if (seq_id > max) {
max = seq_id;
}
if (seq_id < min) {
min = seq_id;
}
// Assuming the tokens are in-order
if (batch.pos[i] != cache.cells[seq_id].pos + 1) {
// What should happen when the pos backtracks or skips a value?
// Clearing the state mid-batch would require special-casing which isn't done.
LLAMA_LOG_WARN("%s: non-consecutive token position %d after %d for sequence %d\n",
__func__, batch.pos[i], cache.cells[seq_id].pos, seq_id);
}
if (cache.cells[seq_id].pos < 0 && 0 <= batch.pos[i]) {
cache.used += 1;
}
cache.cells[seq_id].pos = batch.pos[i];
// NOTE: seq_ids are not inserted here; they are handled when the input tensors are set
} else {
// too big seq_id
// TODO: would it be possible to resize the KV cache size instead?
LLAMA_LOG_ERROR("%s: seq_id=%d >= kv_size=%d Try using a bigger --parallel value\n", __func__, seq_id, cache.size);
return false;
}
}
}
// allow getting the range of used cells, from head to head + n
cache.head = min;
cache.n = max - min + 1;
// sanity check
return max >= min;
}
// otherwise, one cell per token.
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
if (n_tokens > n_ctx) {
LLAMA_LOG_ERROR("%s: n_tokens=%d > n_ctx=%d\n", __func__, n_tokens, n_ctx);
return false;
}
uint32_t n_tested = 0;
while (true) {
if (cache.head + n_tokens > n_ctx) {
kv cache slot search improvements (#3493) * kv cache slot search improvements * Use n_ctx in kv find slot for consistency * Ensure kv cache head points to a valid slot in llama_decode internal * Add some comments to prevent dumb people (like me) from getting confused.
2023-10-06 16:10:13 +00:00
n_tested += n_ctx - cache.head;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
cache.head = 0;
continue;
}
bool found = true;
for (uint32_t i = 0; i < n_tokens; i++) {
if (cache.cells[cache.head + i].pos >= 0) {
found = false;
cache.head += i + 1;
n_tested += i + 1;
break;
}
}
if (found) {
break;
}
if (n_tested >= n_ctx) {
//LLAMA_LOG_ERROR("%s: failed to find a slot for %d tokens\n", __func__, n_tokens);
return false;
}
}
for (uint32_t i = 0; i < n_tokens; i++) {
cache.cells[cache.head + i].pos = batch.pos[i];
speculative : add tree-based sampling example (#3624) * sampling : one sequence per sampling context ggml-ci * speculative : add tree-based sampling support ggml-ci * speculative : reuse the n_parallel CLI param * speculative : refactor sampling * examples : fix build after sampling refactoring ggml-ci * batched : fix n_seq_id * sampling : fix malloc ggml-ci * swift : fix build ggml-ci * swift : try to fix build ggml-ci * prompts : add assistant.txt * common : add llama_batch_add() and llama_batch_clear() helpers * speculative : minor refactor ggml-ci * minor : comments + rename ggml-ci * speculative : fix off-by-one for n_drafted * speculative : fix the n_drafted fix + p constants
2023-10-18 13:21:57 +00:00
for (int32_t j = 0; j < batch.n_seq_id[i]; j++) {
cache.cells[cache.head + i].seq_id.insert(batch.seq_id[i][j]);
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
}
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
cache.used += n_tokens;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
return true;
}
// find how many cells are currently in use
llama : fix llama_copy_state_data with fragmented KV cache (#5840) The row size of the saved states was based on kv_self.head while it should be based on llama_kv_cache_cell_max. Existing session files should still work. * llama : fix llama_kv_cache_cell_max inability to return 1 I've also changed its return type to uint32_t, because this function is always used to set the value of uint32_t variables, and because the index already has this type. * llama : fix state size calculation Some bytes in the state were unaccounted for in llama_get_state_size. Since the logits reserve so much space, it did not cause problems.
2024-03-03 08:41:55 +00:00
static uint32_t llama_kv_cache_cell_max(const struct llama_kv_cache & cache) {
for (uint32_t i = cache.size; i > 0; --i) {
const llama_kv_cell & cell = cache.cells[i - 1];
if (cell.pos >= 0 && !cell.is_empty()) {
return i;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
}
}
return 0;
}
Extend llama_kv_cache_seq_rm to allow matching any sequence (#3843) * Extend llama_kv_cache_seq_rm to allow matichng any sequence * Replace llama_kv_cache_tokens_rm with llama_kv_cache_clear Use llama_kv_cache_clear for cache clearing Change calls to llama_kv_cache_tokens_rm that want to delete by position to use llama_kv_cache_seq_rm functionality
2023-10-29 17:31:40 +00:00
static void llama_kv_cache_clear(struct llama_kv_cache & cache) {
ggml : move FP16 <-> FP32 code to ggml-impl.h (#3861) * ggml : move FP16 <-> FP32 stuff to ggml-impl.h ggml-ci * tests : fix ARM build * ggml : explicitly initialize deprecated type traits * ggml : add math.h to ggml-impl.h * ggml : remove duplicate static assert macros * ggml : prefix lookup tables with ggml_ ggml-ci * ggml-impl : move extern "C" to start of file
2023-10-30 17:19:15 +00:00
for (int32_t i = 0; i < (int32_t) cache.size; ++i) {
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
cache.cells[i].pos = -1;
cache.cells[i].seq_id.clear();
}
Extend llama_kv_cache_seq_rm to allow matching any sequence (#3843) * Extend llama_kv_cache_seq_rm to allow matichng any sequence * Replace llama_kv_cache_tokens_rm with llama_kv_cache_clear Use llama_kv_cache_clear for cache clearing Change calls to llama_kv_cache_tokens_rm that want to delete by position to use llama_kv_cache_seq_rm functionality
2023-10-29 17:31:40 +00:00
cache.head = 0;
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
cache.used = 0;
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
for (auto & buf : cache.bufs) {
ggml_backend_buffer_clear(buf, 0);
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
}
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
static bool llama_kv_cache_seq_rm(
llama : fix session saving/loading (#3400) * llama : fix session saving/loading * llama : temp fix for clearing "future" tokens from the KV cache * llama : fix handling of "future" tokens when loading sessions * llama : fix comments for llama_kv_cache API
2023-10-03 18:04:01 +00:00
struct llama_kv_cache & cache,
llama_seq_id seq_id,
llama_pos p0,
llama_pos p1) {
kv cache slot search improvements (#3493) * kv cache slot search improvements * Use n_ctx in kv find slot for consistency * Ensure kv cache head points to a valid slot in llama_decode internal * Add some comments to prevent dumb people (like me) from getting confused.
2023-10-06 16:10:13 +00:00
uint32_t new_head = cache.size;
llama : fix session saving/loading (#3400) * llama : fix session saving/loading * llama : temp fix for clearing "future" tokens from the KV cache * llama : fix handling of "future" tokens when loading sessions * llama : fix comments for llama_kv_cache API
2023-10-03 18:04:01 +00:00
if (p0 < 0) p0 = 0;
if (p1 < 0) p1 = std::numeric_limits<llama_pos>::max();
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
// models like Mamba can't have a state partially erased
if (cache.recurrent) {
if (seq_id >= (int64_t) cache.size) {
// could be fatal
return false;
}
if (0 <= seq_id) {
// partial intersection is invalid
if ((0 < p0 && p0 <= cache.cells[seq_id].pos) || (0 < p1 && p1 <= cache.cells[seq_id].pos)) {
return false;
}
} else {
// seq_id is negative, then the range should include everything or nothing
if (p0 != p1 && (p0 != 0 || p1 != std::numeric_limits<llama_pos>::max())) {
return false;
}
}
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
for (uint32_t i = 0; i < cache.size; ++i) {
Extend llama_kv_cache_seq_rm to allow matching any sequence (#3843) * Extend llama_kv_cache_seq_rm to allow matichng any sequence * Replace llama_kv_cache_tokens_rm with llama_kv_cache_clear Use llama_kv_cache_clear for cache clearing Change calls to llama_kv_cache_tokens_rm that want to delete by position to use llama_kv_cache_seq_rm functionality
2023-10-29 17:31:40 +00:00
if (cache.cells[i].pos >= p0 && cache.cells[i].pos < p1) {
if (seq_id < 0) {
cache.cells[i].seq_id.clear();
} else if (cache.cells[i].has_seq_id(seq_id)) {
cache.cells[i].seq_id.erase(seq_id);
} else {
continue;
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
if (cache.cells[i].is_empty()) {
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
// keep count of the number of used cells
if (cache.cells[i].pos >= 0) cache.used--;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
cache.cells[i].pos = -1;
kv cache slot search improvements (#3493) * kv cache slot search improvements * Use n_ctx in kv find slot for consistency * Ensure kv cache head points to a valid slot in llama_decode internal * Add some comments to prevent dumb people (like me) from getting confused.
2023-10-06 16:10:13 +00:00
if (new_head == cache.size) new_head = i;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
}
}
}
kv cache slot search improvements (#3493) * kv cache slot search improvements * Use n_ctx in kv find slot for consistency * Ensure kv cache head points to a valid slot in llama_decode internal * Add some comments to prevent dumb people (like me) from getting confused.
2023-10-06 16:10:13 +00:00
// If we freed up a slot, set head to it so searching can start there.
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
if (new_head != cache.size && new_head < cache.head) cache.head = new_head;
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
return true;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
}
static void llama_kv_cache_seq_cp(
llama : fix session saving/loading (#3400) * llama : fix session saving/loading * llama : temp fix for clearing "future" tokens from the KV cache * llama : fix handling of "future" tokens when loading sessions * llama : fix comments for llama_kv_cache API
2023-10-03 18:04:01 +00:00
struct llama_kv_cache & cache,
llama_seq_id seq_id_src,
llama_seq_id seq_id_dst,
llama_pos p0,
llama_pos p1) {
if (p0 < 0) p0 = 0;
if (p1 < 0) p1 = std::numeric_limits<llama_pos>::max();
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
if (cache.recurrent) {
if ((uint32_t) seq_id_dst < cache.size && (uint32_t) seq_id_src < cache.size) {
seq_id_src = cache.cells[seq_id_src].src;
GGML_ASSERT((uint32_t) seq_id_src < cache.size);
// intent to "copy from"
// supports copy chains thanks to taking the source of the source
cache.cells[seq_id_dst].src = seq_id_src;
// preserve the "keep or clear" status of the copied sequence
if (cache.cells[seq_id_src].has_seq_id(seq_id_src)) {
cache.cells[seq_id_dst].seq_id.insert(seq_id_dst);
} else {
cache.cells[seq_id_dst].seq_id.erase(seq_id_dst);
}
cache.do_copy = true;
cache.cells[seq_id_dst].pos = cache.cells[seq_id_src].pos;
}
return;
}
// otherwise, this is the KV cache of a Transformer-like model
kv cache slot search improvements (#3493) * kv cache slot search improvements * Use n_ctx in kv find slot for consistency * Ensure kv cache head points to a valid slot in llama_decode internal * Add some comments to prevent dumb people (like me) from getting confused.
2023-10-06 16:10:13 +00:00
cache.head = 0;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
for (uint32_t i = 0; i < cache.size; ++i) {
if (cache.cells[i].has_seq_id(seq_id_src) && cache.cells[i].pos >= p0 && cache.cells[i].pos < p1) {
cache.cells[i].seq_id.insert(seq_id_dst);
}
}
}
static void llama_kv_cache_seq_keep(struct llama_kv_cache & cache, llama_seq_id seq_id) {
kv cache slot search improvements (#3493) * kv cache slot search improvements * Use n_ctx in kv find slot for consistency * Ensure kv cache head points to a valid slot in llama_decode internal * Add some comments to prevent dumb people (like me) from getting confused.
2023-10-06 16:10:13 +00:00
uint32_t new_head = cache.size;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
for (uint32_t i = 0; i < cache.size; ++i) {
if (!cache.cells[i].has_seq_id(seq_id)) {
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
if (cache.cells[i].pos >= 0) cache.used--;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
cache.cells[i].pos = -1;
cache.cells[i].seq_id.clear();
kv cache slot search improvements (#3493) * kv cache slot search improvements * Use n_ctx in kv find slot for consistency * Ensure kv cache head points to a valid slot in llama_decode internal * Add some comments to prevent dumb people (like me) from getting confused.
2023-10-06 16:10:13 +00:00
if (new_head == cache.size) new_head = i;
speculative : add tree-based sampling example (#3624) * sampling : one sequence per sampling context ggml-ci * speculative : add tree-based sampling support ggml-ci * speculative : reuse the n_parallel CLI param * speculative : refactor sampling * examples : fix build after sampling refactoring ggml-ci * batched : fix n_seq_id * sampling : fix malloc ggml-ci * swift : fix build ggml-ci * swift : try to fix build ggml-ci * prompts : add assistant.txt * common : add llama_batch_add() and llama_batch_clear() helpers * speculative : minor refactor ggml-ci * minor : comments + rename ggml-ci * speculative : fix off-by-one for n_drafted * speculative : fix the n_drafted fix + p constants
2023-10-18 13:21:57 +00:00
} else {
cache.cells[i].seq_id.clear();
cache.cells[i].seq_id.insert(seq_id);
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
}
}
kv cache slot search improvements (#3493) * kv cache slot search improvements * Use n_ctx in kv find slot for consistency * Ensure kv cache head points to a valid slot in llama_decode internal * Add some comments to prevent dumb people (like me) from getting confused.
2023-10-06 16:10:13 +00:00
// If we freed up a slot, set head to it so searching can start there.
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
if (new_head != cache.size && new_head < cache.head) cache.head = new_head;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
static void llama_kv_cache_seq_add(
llama : fix session saving/loading (#3400) * llama : fix session saving/loading * llama : temp fix for clearing "future" tokens from the KV cache * llama : fix handling of "future" tokens when loading sessions * llama : fix comments for llama_kv_cache API
2023-10-03 18:04:01 +00:00
struct llama_kv_cache & cache,
llama_seq_id seq_id,
llama_pos p0,
llama_pos p1,
llama_pos delta) {
kv cache slot search improvements (#3493) * kv cache slot search improvements * Use n_ctx in kv find slot for consistency * Ensure kv cache head points to a valid slot in llama_decode internal * Add some comments to prevent dumb people (like me) from getting confused.
2023-10-06 16:10:13 +00:00
uint32_t new_head = cache.size;
llama : fix session saving/loading (#3400) * llama : fix session saving/loading * llama : temp fix for clearing "future" tokens from the KV cache * llama : fix handling of "future" tokens when loading sessions * llama : fix comments for llama_kv_cache API
2023-10-03 18:04:01 +00:00
if (p0 < 0) p0 = 0;
if (p1 < 0) p1 = std::numeric_limits<llama_pos>::max();
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
if (cache.recurrent) {
// for Mamba-like models, only the pos needs to be shifted
if (0 <= seq_id && seq_id < (int64_t) cache.size) {
llama_kv_cell & cell = cache.cells[seq_id];
if (cell.has_seq_id(seq_id) && p0 <= cell.pos && cell.pos < p1) {
cell.pos += delta;
}
}
return;
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
for (uint32_t i = 0; i < cache.size; ++i) {
if (cache.cells[i].has_seq_id(seq_id) && cache.cells[i].pos >= p0 && cache.cells[i].pos < p1) {
llama : fix kv shift bug (#3835) ggml-ci
2023-10-29 16:32:51 +00:00
cache.has_shift = true;
cache.cells[i].pos += delta;
cache.cells[i].delta += delta;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
if (cache.cells[i].pos < 0) {
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
if (!cache.cells[i].is_empty()) {
cache.used--;
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
cache.cells[i].pos = -1;
cache.cells[i].seq_id.clear();
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
if (new_head == cache.size) {
new_head = i;
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
}
}
}
kv cache slot search improvements (#3493) * kv cache slot search improvements * Use n_ctx in kv find slot for consistency * Ensure kv cache head points to a valid slot in llama_decode internal * Add some comments to prevent dumb people (like me) from getting confused.
2023-10-06 16:10:13 +00:00
// If we freed up a slot, set head to it so searching can start there.
// Otherwise we just start the next search from the beginning.
cache.head = new_head != cache.size ? new_head : 0;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
}
examples : add passkey test (#3856) * examples : add passkey test * passkey : better prints * passkey : select pass key pos from CLI * passkey : simplify n_past logic * make : add passkey target * passkey : add "self-extend"-like context extension (#4810) * llama : "self-extend"-like context extension * passkey : add comment * passkey : add readme
2024-01-08 09:14:04 +00:00
static void llama_kv_cache_seq_div(
struct llama_kv_cache & cache,
llama_seq_id seq_id,
llama_pos p0,
llama_pos p1,
int d) {
if (p0 < 0) p0 = 0;
if (p1 < 0) p1 = std::numeric_limits<llama_pos>::max();
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
if (cache.recurrent) {
// for Mamba-like models, only the pos needs to be changed
if (0 <= seq_id && seq_id < (int64_t) cache.size) {
llama_kv_cell & cell = cache.cells[seq_id];
if (cell.has_seq_id(seq_id) && p0 <= cell.pos && cell.pos < p1) {
cell.pos /= d;
}
}
return;
}
examples : add passkey test (#3856) * examples : add passkey test * passkey : better prints * passkey : select pass key pos from CLI * passkey : simplify n_past logic * make : add passkey target * passkey : add "self-extend"-like context extension (#4810) * llama : "self-extend"-like context extension * passkey : add comment * passkey : add readme
2024-01-08 09:14:04 +00:00
for (uint32_t i = 0; i < cache.size; ++i) {
if (cache.cells[i].has_seq_id(seq_id) && cache.cells[i].pos >= p0 && cache.cells[i].pos < p1) {
cache.has_shift = true;
{
llama_pos p_old = cache.cells[i].pos;
cache.cells[i].pos /= d;
cache.cells[i].delta += cache.cells[i].pos - p_old;
}
}
}
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
static llama_pos llama_kv_cache_seq_pos_max(struct llama_kv_cache & cache, llama_seq_id seq_id) {
llama_pos result = 0;
for (uint32_t i = 0; i < cache.size; ++i) {
if (cache.cells[i].has_seq_id(seq_id)) {
result = std::max(result, cache.cells[i].pos);
}
}
return result;
}
static void llama_kv_cache_defrag(struct llama_kv_cache & cache) {
cache.do_defrag = true;
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
//
// model loading and saving
//
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
enum llama_fver {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
GGUF_FILE_VERSION_V1 = 1,
gguf : add 64-bit support (GGUF v2) (#2821) * gguf : bump version to 2 * gguf : add support for 64-bit (no backwards comp yet) * gguf : v1 backwards comp * gguf.py : bump GGUF version * gguf.py : uint64_t on all lengths, sizes and counts, enums still uint32_t * gguf.py : string lengths uint32_t * gguf : update all counts to 64-bit * gguf.py : string len uint64_t and n_dims uint32_t * gguf : fix typo * llama.cpp : print gguf version --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com>
2023-08-27 11:19:54 +00:00
GGUF_FILE_VERSION_V2 = 2,
llama : correctly report GGUFv3 format (#3818)
2023-10-27 21:33:53 +00:00
GGUF_FILE_VERSION_V3 = 3,
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
};
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
static const char * llama_file_version_name(llama_fver version) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
switch (version) {
gguf : add 64-bit support (GGUF v2) (#2821) * gguf : bump version to 2 * gguf : add support for 64-bit (no backwards comp yet) * gguf : v1 backwards comp * gguf.py : bump GGUF version * gguf.py : uint64_t on all lengths, sizes and counts, enums still uint32_t * gguf.py : string lengths uint32_t * gguf : update all counts to 64-bit * gguf.py : string len uint64_t and n_dims uint32_t * gguf : fix typo * llama.cpp : print gguf version --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com>
2023-08-27 11:19:54 +00:00
case GGUF_FILE_VERSION_V1: return "GGUF V1 (support until nov 2023)";
llama : correctly report GGUFv3 format (#3818)
2023-10-27 21:33:53 +00:00
case GGUF_FILE_VERSION_V2: return "GGUF V2";
case GGUF_FILE_VERSION_V3: return "GGUF V3 (latest)";
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
return "unknown";
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
static std::string llama_format_tensor_shape(const std::vector<int64_t> & ne) {
stdout : vertical align outputs for better readibility
2023-04-16 10:58:48 +00:00
char buf[256];
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
snprintf(buf, sizeof(buf), "%5" PRId64, ne.at(0));
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
for (size_t i = 1; i < ne.size(); i++) {
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
snprintf(buf + strlen(buf), sizeof(buf) - strlen(buf), ", %5" PRId64, ne.at(i));
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
stdout : vertical align outputs for better readibility
2023-04-16 10:58:48 +00:00
return buf;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
static std::string llama_format_tensor_shape(const struct ggml_tensor * t) {
char buf[256];
snprintf(buf, sizeof(buf), "%5" PRId64, t->ne[0]);
for (int i = 1; i < GGML_MAX_DIMS; i++) {
snprintf(buf + strlen(buf), sizeof(buf) - strlen(buf), ", %5" PRId64, t->ne[i]);
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
return buf;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
namespace GGUFMeta {
template <typename T, gguf_type gt_, T (*gfun)(const gguf_context *, const int)>
struct GKV_Base_Type {
static constexpr gguf_type gt = gt_;
static T getter(const gguf_context * ctx, const int kid) {
return gfun(ctx, kid);
}
};
template<typename T> struct GKV_Base;
template<> struct GKV_Base<bool >: GKV_Base_Type<bool, GGUF_TYPE_BOOL, gguf_get_val_bool> {};
template<> struct GKV_Base<uint8_t >: GKV_Base_Type<uint8_t, GGUF_TYPE_UINT8, gguf_get_val_u8 > {};
template<> struct GKV_Base<uint16_t >: GKV_Base_Type<uint16_t, GGUF_TYPE_UINT16, gguf_get_val_u16 > {};
template<> struct GKV_Base<uint32_t >: GKV_Base_Type<uint32_t, GGUF_TYPE_UINT32, gguf_get_val_u32 > {};
template<> struct GKV_Base<uint64_t >: GKV_Base_Type<uint64_t, GGUF_TYPE_UINT64, gguf_get_val_u64 > {};
template<> struct GKV_Base<int8_t >: GKV_Base_Type<int8_t, GGUF_TYPE_INT8, gguf_get_val_i8 > {};
template<> struct GKV_Base<int16_t >: GKV_Base_Type<int16_t, GGUF_TYPE_INT16, gguf_get_val_i16 > {};
template<> struct GKV_Base<int32_t >: GKV_Base_Type<int32_t, GGUF_TYPE_INT32, gguf_get_val_i32 > {};
template<> struct GKV_Base<int64_t >: GKV_Base_Type<int64_t, GGUF_TYPE_INT64, gguf_get_val_i64 > {};
template<> struct GKV_Base<float >: GKV_Base_Type<float, GGUF_TYPE_FLOAT32, gguf_get_val_f32 > {};
template<> struct GKV_Base<double >: GKV_Base_Type<double, GGUF_TYPE_FLOAT64, gguf_get_val_f64 > {};
template<> struct GKV_Base<const char *>: GKV_Base_Type<const char *, GGUF_TYPE_STRING, gguf_get_val_str > {};
template<> struct GKV_Base<std::string> {
static constexpr gguf_type gt = GGUF_TYPE_STRING;
static std::string getter(const gguf_context * ctx, const int kid) {
return gguf_get_val_str(ctx, kid);
}
};
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
struct ArrayInfo {
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
const gguf_type gt;
const size_t length;
const void * data;
};
template<> struct GKV_Base<ArrayInfo> {
public:
static constexpr gguf_type gt = GGUF_TYPE_ARRAY;
static ArrayInfo getter(const gguf_context *ctx, const int k) {
return ArrayInfo {
gguf_get_arr_type(ctx, k),
size_t(gguf_get_arr_n(ctx, k)),
gguf_get_arr_data(ctx, k),
};
}
};
template<typename T>
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
class GKV : public GKV_Base<T> {
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
GKV() = delete;
public:
static T get_kv(const gguf_context * ctx, const int k) {
const enum gguf_type kt = gguf_get_kv_type(ctx, k);
if (kt != GKV::gt) {
throw std::runtime_error(format("key %s has wrong type %s but expected type %s",
gguf_get_key(ctx, k), gguf_type_name(kt), gguf_type_name(GKV::gt)));
}
return GKV::getter(ctx, k);
}
static const char * override_type_to_str(const llama_model_kv_override_type ty) {
switch (ty) {
code : normalize enum names (#5697) * coda : normalize enum names ggml-ci * code : cont * code : cont
2024-02-25 10:09:09 +00:00
case LLAMA_KV_OVERRIDE_TYPE_BOOL: return "bool";
case LLAMA_KV_OVERRIDE_TYPE_INT: return "int";
case LLAMA_KV_OVERRIDE_TYPE_FLOAT: return "float";
quantize: add imatrix and dataset metadata in GGUF (#6658) * imatrix: save the dataset file used in the output file * llama: support kv overrides type string string * common: factorize KV Overrides parsing between common and server * quantize: add imatrix n entries and dataset KV metadata quantize: factorize KV Overrides parsing between common #6656 * llama: remove kv override str_value initialization as it does not compile on some toolchain * quantize: add imatrix m_last_call as `quantize.imatrix.chunks_count` * quantize: add imatrix filename in KV * llama: add llama_model_kv_override_free * common: add llama_model_kv_override_free common: free kv override if used after model loading * llama: finally move the string KV override value to the stack * llama : minor * no need to add a NUL to the std::vector, std::string can be initialized from a pair of iterators. Co-authored-by: slaren <slarengh@gmail.com> * kv override: ensure string termination --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-04-26 18:06:33 +00:00
case LLAMA_KV_OVERRIDE_TYPE_STR: return "str";
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
}
return "unknown";
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
static bool validate_override(const llama_model_kv_override_type expected_type, const struct llama_model_kv_override * ovrd) {
if (!ovrd) { return false; }
if (ovrd->tag == expected_type) {
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
LLAMA_LOG_INFO("%s: Using metadata override (%5s) '%s' = ",
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
__func__, override_type_to_str(ovrd->tag), ovrd->key);
switch (ovrd->tag) {
code : normalize enum names (#5697) * coda : normalize enum names ggml-ci * code : cont * code : cont
2024-02-25 10:09:09 +00:00
case LLAMA_KV_OVERRIDE_TYPE_BOOL: {
quantize: add imatrix and dataset metadata in GGUF (#6658) * imatrix: save the dataset file used in the output file * llama: support kv overrides type string string * common: factorize KV Overrides parsing between common and server * quantize: add imatrix n entries and dataset KV metadata quantize: factorize KV Overrides parsing between common #6656 * llama: remove kv override str_value initialization as it does not compile on some toolchain * quantize: add imatrix m_last_call as `quantize.imatrix.chunks_count` * quantize: add imatrix filename in KV * llama: add llama_model_kv_override_free * common: add llama_model_kv_override_free common: free kv override if used after model loading * llama: finally move the string KV override value to the stack * llama : minor * no need to add a NUL to the std::vector, std::string can be initialized from a pair of iterators. Co-authored-by: slaren <slarengh@gmail.com> * kv override: ensure string termination --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-04-26 18:06:33 +00:00
LLAMA_LOG_INFO("%s\n", ovrd->val_bool ? "true" : "false");
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
} break;
code : normalize enum names (#5697) * coda : normalize enum names ggml-ci * code : cont * code : cont
2024-02-25 10:09:09 +00:00
case LLAMA_KV_OVERRIDE_TYPE_INT: {
quantize: add imatrix and dataset metadata in GGUF (#6658) * imatrix: save the dataset file used in the output file * llama: support kv overrides type string string * common: factorize KV Overrides parsing between common and server * quantize: add imatrix n entries and dataset KV metadata quantize: factorize KV Overrides parsing between common #6656 * llama: remove kv override str_value initialization as it does not compile on some toolchain * quantize: add imatrix m_last_call as `quantize.imatrix.chunks_count` * quantize: add imatrix filename in KV * llama: add llama_model_kv_override_free * common: add llama_model_kv_override_free common: free kv override if used after model loading * llama: finally move the string KV override value to the stack * llama : minor * no need to add a NUL to the std::vector, std::string can be initialized from a pair of iterators. Co-authored-by: slaren <slarengh@gmail.com> * kv override: ensure string termination --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-04-26 18:06:33 +00:00
LLAMA_LOG_INFO("%" PRId64 "\n", ovrd->val_i64);
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
} break;
code : normalize enum names (#5697) * coda : normalize enum names ggml-ci * code : cont * code : cont
2024-02-25 10:09:09 +00:00
case LLAMA_KV_OVERRIDE_TYPE_FLOAT: {
quantize: add imatrix and dataset metadata in GGUF (#6658) * imatrix: save the dataset file used in the output file * llama: support kv overrides type string string * common: factorize KV Overrides parsing between common and server * quantize: add imatrix n entries and dataset KV metadata quantize: factorize KV Overrides parsing between common #6656 * llama: remove kv override str_value initialization as it does not compile on some toolchain * quantize: add imatrix m_last_call as `quantize.imatrix.chunks_count` * quantize: add imatrix filename in KV * llama: add llama_model_kv_override_free * common: add llama_model_kv_override_free common: free kv override if used after model loading * llama: finally move the string KV override value to the stack * llama : minor * no need to add a NUL to the std::vector, std::string can be initialized from a pair of iterators. Co-authored-by: slaren <slarengh@gmail.com> * kv override: ensure string termination --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-04-26 18:06:33 +00:00
LLAMA_LOG_INFO("%.6f\n", ovrd->val_f64);
} break;
case LLAMA_KV_OVERRIDE_TYPE_STR: {
LLAMA_LOG_INFO("%s\n", ovrd->val_str);
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
} break;
default:
// Shouldn't be possible to end up here, but just in case...
throw std::runtime_error(
format("Unsupported attempt to override %s type for metadata key %s\n",
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
override_type_to_str(ovrd->tag), ovrd->key));
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
}
return true;
}
LLAMA_LOG_WARN("%s: Warning: Bad metadata override type for key '%s', expected %s but got %s\n",
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
__func__, ovrd->key, override_type_to_str(expected_type), override_type_to_str(ovrd->tag));
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
return false;
}
template<typename OT>
static typename std::enable_if<std::is_same<OT, bool>::value, bool>::type
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
try_override(OT & target, const struct llama_model_kv_override * ovrd) {
if (validate_override(LLAMA_KV_OVERRIDE_TYPE_BOOL, ovrd)) {
quantize: add imatrix and dataset metadata in GGUF (#6658) * imatrix: save the dataset file used in the output file * llama: support kv overrides type string string * common: factorize KV Overrides parsing between common and server * quantize: add imatrix n entries and dataset KV metadata quantize: factorize KV Overrides parsing between common #6656 * llama: remove kv override str_value initialization as it does not compile on some toolchain * quantize: add imatrix m_last_call as `quantize.imatrix.chunks_count` * quantize: add imatrix filename in KV * llama: add llama_model_kv_override_free * common: add llama_model_kv_override_free common: free kv override if used after model loading * llama: finally move the string KV override value to the stack * llama : minor * no need to add a NUL to the std::vector, std::string can be initialized from a pair of iterators. Co-authored-by: slaren <slarengh@gmail.com> * kv override: ensure string termination --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-04-26 18:06:33 +00:00
target = ovrd->val_bool;
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
return true;
}
llama : fix try_override for bool_value which always return true (#4519)
2023-12-18 13:14:58 +00:00
return false;
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
}
template<typename OT>
static typename std::enable_if<!std::is_same<OT, bool>::value && std::is_integral<OT>::value, bool>::type
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
try_override(OT & target, const struct llama_model_kv_override * ovrd) {
if (validate_override(LLAMA_KV_OVERRIDE_TYPE_INT, ovrd)) {
quantize: add imatrix and dataset metadata in GGUF (#6658) * imatrix: save the dataset file used in the output file * llama: support kv overrides type string string * common: factorize KV Overrides parsing between common and server * quantize: add imatrix n entries and dataset KV metadata quantize: factorize KV Overrides parsing between common #6656 * llama: remove kv override str_value initialization as it does not compile on some toolchain * quantize: add imatrix m_last_call as `quantize.imatrix.chunks_count` * quantize: add imatrix filename in KV * llama: add llama_model_kv_override_free * common: add llama_model_kv_override_free common: free kv override if used after model loading * llama: finally move the string KV override value to the stack * llama : minor * no need to add a NUL to the std::vector, std::string can be initialized from a pair of iterators. Co-authored-by: slaren <slarengh@gmail.com> * kv override: ensure string termination --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-04-26 18:06:33 +00:00
target = ovrd->val_i64;
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
return true;
}
return false;
}
template<typename OT>
static typename std::enable_if<std::is_floating_point<OT>::value, bool>::type
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
try_override(T & target, const struct llama_model_kv_override * ovrd) {
if (validate_override(LLAMA_KV_OVERRIDE_TYPE_FLOAT, ovrd)) {
quantize: add imatrix and dataset metadata in GGUF (#6658) * imatrix: save the dataset file used in the output file * llama: support kv overrides type string string * common: factorize KV Overrides parsing between common and server * quantize: add imatrix n entries and dataset KV metadata quantize: factorize KV Overrides parsing between common #6656 * llama: remove kv override str_value initialization as it does not compile on some toolchain * quantize: add imatrix m_last_call as `quantize.imatrix.chunks_count` * quantize: add imatrix filename in KV * llama: add llama_model_kv_override_free * common: add llama_model_kv_override_free common: free kv override if used after model loading * llama: finally move the string KV override value to the stack * llama : minor * no need to add a NUL to the std::vector, std::string can be initialized from a pair of iterators. Co-authored-by: slaren <slarengh@gmail.com> * kv override: ensure string termination --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-04-26 18:06:33 +00:00
target = ovrd->val_f64;
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
return true;
}
return false;
}
template<typename OT>
static typename std::enable_if<std::is_same<OT, std::string>::value, bool>::type
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
try_override(T & target, const struct llama_model_kv_override * ovrd) {
quantize: add imatrix and dataset metadata in GGUF (#6658) * imatrix: save the dataset file used in the output file * llama: support kv overrides type string string * common: factorize KV Overrides parsing between common and server * quantize: add imatrix n entries and dataset KV metadata quantize: factorize KV Overrides parsing between common #6656 * llama: remove kv override str_value initialization as it does not compile on some toolchain * quantize: add imatrix m_last_call as `quantize.imatrix.chunks_count` * quantize: add imatrix filename in KV * llama: add llama_model_kv_override_free * common: add llama_model_kv_override_free common: free kv override if used after model loading * llama: finally move the string KV override value to the stack * llama : minor * no need to add a NUL to the std::vector, std::string can be initialized from a pair of iterators. Co-authored-by: slaren <slarengh@gmail.com> * kv override: ensure string termination --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-04-26 18:06:33 +00:00
if (validate_override(LLAMA_KV_OVERRIDE_TYPE_STR, ovrd)) {
target = ovrd->val_str;
return true;
}
return false;
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
static bool set(const gguf_context * ctx, const int k, T & target, const struct llama_model_kv_override * ovrd = nullptr) {
if (try_override<T>(target, ovrd)) {
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
return true;
}
if (k < 0) { return false; }
target = get_kv(ctx, k);
return true;
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
static bool set(const gguf_context * ctx, const char * key, T & target, const struct llama_model_kv_override * ovrd = nullptr) {
return set(ctx, gguf_find_key(ctx, key), target, ovrd);
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
static bool set(const gguf_context * ctx, const std::string & key, T & target, const struct llama_model_kv_override * ovrd = nullptr) {
return set(ctx, key.c_str(), target, ovrd);
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
}
};
}
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
using llama_buf_map = std::unordered_map<uint32_t, ggml_backend_buffer_t>;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
struct llama_model_loader {
int n_kv = 0;
int n_tensors = 0;
int n_created = 0;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
int64_t n_elements = 0;
llama.cpp : show model size and BPW on load (#3223)
2023-09-17 12:33:28 +00:00
size_t n_bytes = 0;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
bool use_mmap = false;
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
bool check_tensors;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
llama_files files;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama_ftype ftype;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama_fver fver;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
llama_mmaps mappings;
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
// Holds information on a model weight
struct llama_tensor_weight {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
uint16_t idx; // source file index
size_t offs; // tensor data offset in the original file
ggml_tensor * tensor;
llama : check that all the tensor data is in the model file (#6885) * llama : check that all the tensor data is in the model file * also check for unsigned overflow
2024-04-25 13:23:47 +00:00
llama_tensor_weight(const llama_file * file, uint16_t idx, const char * name, const struct gguf_context * gguf_ctx, ggml_tensor * tensor) : idx(idx), tensor(tensor) {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
const int tensor_idx = gguf_find_tensor(gguf_ctx, name);
offs = gguf_get_data_offset(gguf_ctx) + gguf_get_tensor_offset(gguf_ctx, tensor_idx);
llama : check that all the tensor data is in the model file (#6885) * llama : check that all the tensor data is in the model file * also check for unsigned overflow
2024-04-25 13:23:47 +00:00
if (offs + ggml_nbytes(tensor) < offs || offs + ggml_nbytes(tensor) > file->size) {
throw std::runtime_error(format("tensor '%s' data is not within the file bounds, model is corrupted or incomplete", name));
}
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
}
};
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
std::vector<llama_tensor_weight> weights;
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
std::unordered_map<std::string, struct llama_model_kv_override> kv_overrides;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
struct gguf_context * meta = NULL;
std::vector<ggml_context *> contexts;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
std::string arch_name;
LLM_KV llm_kv = LLM_KV(LLM_ARCH_UNKNOWN);
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
llama_model_loader(const std::string & fname, bool use_mmap, bool check_tensors, const struct llama_model_kv_override * param_overrides_p) {
llama : check LLAMA_TRACE env for extra logging (#4929) * llama : minor fix indent * llama : check LLAMA_TRACE env for extra logging ggml-ci
2024-01-14 11:26:53 +00:00
int trace = 0;
if (getenv("LLAMA_TRACE")) {
trace = atoi(getenv("LLAMA_TRACE"));
}
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
if (param_overrides_p != nullptr) {
for (const struct llama_model_kv_override *p = param_overrides_p; p->key[0] != 0; p++) {
kv_overrides.insert({std::string(p->key), *p});
}
}
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
struct ggml_context * ctx = NULL;
struct gguf_init_params params = {
/*.no_alloc = */ true,
/*.ctx = */ &ctx,
};
meta = gguf_init_from_file(fname.c_str(), params);
if (!meta) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
throw std::runtime_error(format("%s: failed to load model from %s\n", __func__, fname.c_str()));
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
get_key(llm_kv(LLM_KV_GENERAL_ARCHITECTURE), arch_name, false);
llm_kv = LLM_KV(llm_arch_from_string(arch_name));
llama : check that all the tensor data is in the model file (#6885) * llama : check that all the tensor data is in the model file * also check for unsigned overflow
2024-04-25 13:23:47 +00:00
files.emplace_back(new llama_file(fname.c_str(), "rb"));
contexts.emplace_back(ctx);
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
// Save tensors data offset of the main file.
// For subsidiary files, `meta` tensor data offset must not be used,
// so we build a unified tensors index for weights.
for (ggml_tensor * cur = ggml_get_first_tensor(ctx); cur; cur = ggml_get_next_tensor(ctx, cur)) {
llama : check that all the tensor data is in the model file (#6885) * llama : check that all the tensor data is in the model file * also check for unsigned overflow
2024-04-25 13:23:47 +00:00
weights.emplace_back(files.back().get(), 0, cur->name, meta, cur);
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
}
uint16_t n_split = 0;
get_key(llm_kv(LLM_KV_SPLIT_COUNT), n_split, false);
// Load additional GGML contexts
if (n_split > 1) {
uint16_t idx = 0;
get_key(llm_kv(LLM_KV_SPLIT_NO), idx);
if (idx != 0) {
throw std::runtime_error(format("illegal split file: %d, model must be loaded with the first split", idx));
}
char split_prefix[PATH_MAX] = {0};
if (!llama_split_prefix(split_prefix, sizeof(split_prefix), fname.c_str(), idx, n_split)) {
throw std::runtime_error(format("invalid split file: %s", fname.c_str()));
}
llama : define magic numbers as integer constants (#1518) (#1520) The underlying representation of multibyte character literals is implementation-defined. This could, at least in principle, cause cross-build data export/import issues independent of endianness. Define magic numbers as integer literals to be on the safe side. Signed-off-by: Juuso Alasuutari <juuso.alasuutari@gmail.com>
2023-05-20 12:58:15 +00:00
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
if (trace > 0) {
LLAMA_LOG_INFO("%s: loading additional %d GGUFs\n", __func__, n_split);
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
char split_path[PATH_MAX] = {0};
for (idx = 1; idx < n_split; idx++) {
llama_split_path(split_path, sizeof(split_path), split_prefix, idx, n_split);
struct gguf_init_params split_params = {
/*.no_alloc = */ true,
/*.ctx = */ &ctx,
};
struct gguf_context * ctx_gguf = gguf_init_from_file(split_path, split_params);
if (!ctx_gguf) {
throw std::runtime_error(format("%s: failed to load GGUF split from %s\n", __func__, split_path));
}
llama : check that all the tensor data is in the model file (#6885) * llama : check that all the tensor data is in the model file * also check for unsigned overflow
2024-04-25 13:23:47 +00:00
files.emplace_back(new llama_file(split_path, "rb"));
contexts.emplace_back(ctx);
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
// Save tensors data offset info of the shard.
for (ggml_tensor * cur = ggml_get_first_tensor(ctx); cur; cur = ggml_get_next_tensor(ctx, cur)) {
llama : check that all the tensor data is in the model file (#6885) * llama : check that all the tensor data is in the model file * also check for unsigned overflow
2024-04-25 13:23:47 +00:00
weights.emplace_back(files.back().get(), idx, cur->name, ctx_gguf, cur);
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
}
gguf_free(ctx_gguf);
}
get_key(llm_kv(LLM_KV_SPLIT_TENSORS_COUNT), n_tensors);
// sanity check
{
const int n_tensors_loaded = (int) weights.size();
if (n_tensors != n_tensors_loaded) {
throw std::runtime_error(format("corrupted model: %d tensors expected but %d found", n_tensors, n_tensors_loaded));
}
}
common: llama_load_model_from_url split support (#6192) * llama: llama_split_prefix fix strncpy does not include string termination common: llama_load_model_from_url: - fix header name case sensitive - support downloading additional split in parallel - hide password in url * common: EOL EOF * common: remove redundant LLAMA_CURL_MAX_PATH_LENGTH definition * common: change max url max length * common: minor comment * server: support HF URL options * llama: llama_model_loader fix log * common: use a constant for max url length * common: clean up curl if file cannot be loaded in gguf * server: tests: add split tests, and HF options params * common: move llama_download_hide_password_in_url inside llama_download_file as a lambda * server: tests: enable back Release test on PR * spacing Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * spacing Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * spacing Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 17:07:00 +00:00
LLAMA_LOG_INFO("%s: additional %d GGUFs metadata loaded.\n", __func__, n_split - 1);
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
}
n_kv = gguf_get_n_kv(meta);
n_tensors = weights.size();
fver = (enum llama_fver) gguf_get_version(meta);
gguf : enforce that tensor names are unique (#6905) * not allow adding duplicated tensor name * no duplicated tensor while reading gguf * typo * throw exception inside llama_model_loader Co-authored-by: slaren <slarengh@gmail.com> --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-04-28 15:36:18 +00:00
std::set<std::string> tensor_names;
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
for (auto & w : weights) {
n_elements += ggml_nelements(w.tensor);
n_bytes += ggml_nbytes(w.tensor);
gguf : enforce that tensor names are unique (#6905) * not allow adding duplicated tensor name * no duplicated tensor while reading gguf * typo * throw exception inside llama_model_loader Co-authored-by: slaren <slarengh@gmail.com> --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-04-28 15:36:18 +00:00
// make sure there is no duplicated tensor names
const std::string name(w.tensor->name);
auto found = tensor_names.find(name);
if (found != tensor_names.end()) {
throw std::runtime_error(format("invalid model: tensor '%s' is duplicated", w.tensor->name));
}
tensor_names.insert(name);
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
llama : define magic numbers as integer constants (#1518) (#1520) The underlying representation of multibyte character literals is implementation-defined. This could, at least in principle, cause cross-build data export/import issues independent of endianness. Define magic numbers as integer literals to be on the safe side. Signed-off-by: Juuso Alasuutari <juuso.alasuutari@gmail.com>
2023-05-20 12:58:15 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
LLAMA_LOG_INFO("%s: loaded meta data with %d key-value pairs and %d tensors from %s (version %s)\n",
__func__, n_kv, n_tensors, fname.c_str(), llama_file_version_name(fver));
llama : grouped-query attention + LLaMAv2 70B support (#2276) * CUDA: GQA implementation * llama : support for GQA and LLaMAv2 70B ggml-ci * py : fix hparams parsing (if-else blocks) ggml-ci * py : oh boy .. ggml-ci * help : fix gqa value for 70B ggml-ci --------- Co-authored-by: JohannesGaessler <johannesg@5d6.de>
2023-07-23 12:09:47 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
// determine file type based on the number of tensors for each quantization and print meta data
// TODO: make optional
{
std::map<enum ggml_type, uint32_t> n_type;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
uint32_t n_type_max = 0;
enum ggml_type type_max = GGML_TYPE_F32;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
for (int i = 0; i < n_tensors; i++) {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
const ggml_tensor * tensor = weights.at(i).tensor;
enum ggml_type type = tensor->type;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
n_type[type]++;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
if (n_type_max < n_type[type]) {
n_type_max = n_type[type];
type_max = type;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : check LLAMA_TRACE env for extra logging (#4929) * llama : minor fix indent * llama : check LLAMA_TRACE env for extra logging ggml-ci
2024-01-14 11:26:53 +00:00
if (trace > 0) {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
const uint16_t sid = weights.at(i).idx;
LLAMA_LOG_INFO("%s: - tensor %4d, split %2d: %32s %-8s [ %s ]\n", __func__, i, sid, ggml_get_name(tensor), ggml_type_name(type), llama_format_tensor_shape(tensor).c_str());
llama : check LLAMA_TRACE env for extra logging (#4929) * llama : minor fix indent * llama : check LLAMA_TRACE env for extra logging ggml-ci
2024-01-14 11:26:53 +00:00
}
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
switch (type_max) {
minor : clean-up some warnings and style (#5094) * minor : clean-up some warnings and style ggml-ci * ggml : add comment
2024-01-23 12:12:57 +00:00
case GGML_TYPE_F32: ftype = LLAMA_FTYPE_ALL_F32; break;
case GGML_TYPE_F16: ftype = LLAMA_FTYPE_MOSTLY_F16; break;
case GGML_TYPE_Q4_0: ftype = LLAMA_FTYPE_MOSTLY_Q4_0; break;
case GGML_TYPE_Q4_1: ftype = LLAMA_FTYPE_MOSTLY_Q4_1; break;
case GGML_TYPE_Q5_0: ftype = LLAMA_FTYPE_MOSTLY_Q5_0; break;
case GGML_TYPE_Q5_1: ftype = LLAMA_FTYPE_MOSTLY_Q5_1; break;
case GGML_TYPE_Q8_0: ftype = LLAMA_FTYPE_MOSTLY_Q8_0; break;
case GGML_TYPE_Q2_K: ftype = LLAMA_FTYPE_MOSTLY_Q2_K; break;
case GGML_TYPE_Q3_K: ftype = LLAMA_FTYPE_MOSTLY_Q3_K_M; break;
case GGML_TYPE_Q4_K: ftype = LLAMA_FTYPE_MOSTLY_Q4_K_M; break;
case GGML_TYPE_Q5_K: ftype = LLAMA_FTYPE_MOSTLY_Q5_K_M; break;
case GGML_TYPE_Q6_K: ftype = LLAMA_FTYPE_MOSTLY_Q6_K; break;
SOTA 2-bit quants (#4773) * iq2_xxs: basics * iq2_xxs: scalar and AVX2 dot products Needed to change Q8_K to have quants in the -127...127 range, else the IQ2_XXS AVX implementation becomes very awkward. The alternative would have been to use Q8_0 instead. Perhaps I'll change later, for now this is what we have. * iq2_xxs: ARM_NEON dot product Somehow strangely slow (112 ms/token). * iq2_xxs: WIP Metal Dequantize works, something is still wrong with the dot product. * iq2_xxs: Metal dot product now works We have PP-512 = 475 t/s TG-128 = 47.3 t/s Not the greatest performance, but not complete garbage either. * iq2_xxs: slighty faster dot product TG-128 is now 48.4 t/s * iq2_xxs: slighty faster dot product TG-128 is now 50.9 t/s * iq2_xxs: even faster Metal dot product TG-128 is now 54.1 t/s. Strangely enough, putting the signs lookup table into shared memory has a bigger impact than the grid values being in shared memory. * iq2_xxs: dequantize CUDA kernel - fix conflict with master * iq2_xxs: quantized CUDA dot product (MMVQ) We get TG-128 = 153.1 t/s * iq2_xxs: slightly faster CUDA dot product TG-128 is now at 155.1 t/s. * iq2_xxs: add to llama ftype enum * iq2_xxs: fix MoE on Metal * Fix missing MMQ ops when on hipBLAS I had put the ggml_supports_mmq call at the wrong place. * Fix bug in qequantize_row_iq2_xxs The 0.25f factor was missing. Great detective work by @ggerganov! * Fixing tests * PR suggestion --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-08 15:02:32 +00:00
case GGML_TYPE_IQ2_XXS: ftype = LLAMA_FTYPE_MOSTLY_IQ2_XXS; break;
ggml : SOTA 2-bit quants (add IQ2_XS) (#4856) * iq2_xs: basics * iq2_xs: this should have been in the basics * iq2_xs: CUDA and scalar CPU works * iq2_xs: WIP Metal * iq2_xs: Metal now works * iq2_xs: working, but dog slow, ARM_NEON dot product * iq2_xs: better ARM_NEON dot product We are now at 19.5 t/s for TG-128 and 61 t/s for PP-512 when running on the CPU. * iq2_xs: AVX2 dot product - 19.5 t/s * iq2_xs: faster AVX2 dit product 21.4 t/s for TG-128, 59.2 t/s for PP-512. The latter is 2x compared to the previous version. * iq2_xs: had forgotten to delete iq2-data.h * Add llama enum for IQ2_XS --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-11 19:39:39 +00:00
case GGML_TYPE_IQ2_XS: ftype = LLAMA_FTYPE_MOSTLY_IQ2_XS; break;
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
case GGML_TYPE_IQ2_S: ftype = LLAMA_FTYPE_MOSTLY_IQ2_S; break;
SOTA 3-bit quants (#5196) * iq3_xxs: quantize/dequantize RMSE seems a bit high-ish at about half-way between q2_K and q3_K, so need to check more. * iq3_xxs: CUDA dequantize works * iq2_xxs: tuning quantization * iq3_xxs: starting to look better PPL on wiki.test.raw LLaMA-v1-7B: 6.4218 LLaMA-v2-7B: 6.3560 Mistral-7B : 6.0717 This is better than Q3_K_XS, with a 5% reduction in quantized model size. * iq3_xxs: CUDA dot product We have PP-512: 5891 t/s TG-128: 143.9 t/s * iq3_xxs: scalar and AVX2 dot products * iq3_xxs: ARM_NEON and Metal Metal performance is decent, ARM_NEON is pathetic * iq3_xxs: slightly better grid points * Faster iq3_xxs and iq2_xs dot products on CUDA * iq3_xxs: add some quant mix * iq3_xxs: fix failing quantization test Dot product still fails. Is this real? * iq3_xxs: hopefully fix ROCm * iq3_xxs: failing tests This time the dot product accuracy did find an actual bug in the AVX2 implementation. * Add IQ3_XXS to test-backend-ops --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-30 13:14:12 +00:00
case GGML_TYPE_IQ3_XXS: ftype = LLAMA_FTYPE_MOSTLY_IQ3_XXS; break;
1.5 bit quantization (#5453) * iq1_s: WIP basics * iq1_s: CUDA is working * iq1_s: scalar CPU dot product * iq1_s: WIP AVX2 dot product - something is not right * Fix tests * Fix shadow warnings * Fix after merge with latest master * iq1_s: AVX2 finally works * iq1_s: ARM_NEON dot product. Works, but not very fast * iq1_s: better grid * iq1_s: use IQ2_XXS for attn_output At a cost of 0.04 extra bpw this gives a big improvement in PPL. * iq1_s: Metal basics Dequantize works, but not dot product * iq1_s: Metal works, but quite slow As usual, Apple Silicon does not like the code I write. * iq1_s: Tests * iq1_s: slightly faster dot product --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-18 16:16:55 +00:00
case GGML_TYPE_IQ1_S: ftype = LLAMA_FTYPE_MOSTLY_IQ1_S; break;
IQ1_M: 1.75 bpw quantization (#6302) * iq1_m: basics * iq1_m: basics-2 * iq1_m: CUDA dequantize works Very 1st shot I get PPL = 9.76 for LLaMA-v2-7B. * iq1_m: separate shifts for each group of 8 in a block We get PPL(LLaMA-v2-7B ) = 9.2810 PPL(LLaMA-v2-13B) = 6.8105 Not bad, but slightly higher than sqrt(PPL(IQ1_S) * PPL(IQ2_XXS)) which is the expected outcome given that IQ1_M is halfway between IQ1_S and IQ2_XXS in terms of bpw. From this, we would expect PPL = 9.14 for LLaMA-v2-7B PPL = 6.63 for LLaMA-v2-13B * iq1_m: go to 3-bit scales There is slight increase in PPL, but the 0.0625 bpw reduction in size is totally worth it. We now have PPL(LLaMA-v2-7B ) = 9.4469 at 1.96 bpw PPL(LLaMA-v2-13B) = 6.8717 at 1.93 bpw PPL(LLaMA-v2-70B) = 4.8568 at 1.85 bpw * iq1_m: scalar dot product * iq1_m: AVX2 dot product * iq1_m: very slightly faster AVX2 dot product * iq1_m: ARM_NEON dot product Works, but very slow (10.5 t/s) * iq1_m: Metal - dequantize works, dot product does not * iq1_m: Metal now works About the same performance as iq1_s. * iq1_m: minor * iq1_m: checking pure iq1_m quantization It is pretty bad: PPL(LLaMA-v2-7B) = 34 if we quantize output.weight with Q4_K. * iiq1_m: slightly faster ARM_NEON dot product 10.5 t/s -> 11.65 t/s * iq1_m: faster ARM_NEON dot product 11.65 t/s -> 14.9 t/s * iq1_m: another minor ARM_NEON dot product improvement 14.9 -> 15.0 t/s * iq1_m: small PPL improvement via super-block scale adjustment After quantizing block scales redo the super-block scale fit. PPL(LLaMA-v2-7B ) = 9.3346 PPL(LLaMA-v2-13B) = 6.8419 PPL(LLaMA-v2-70B) = 4.8294 PPL(Mistral-7B ) = 8.1624 * iq1_m: adapt to CUDA refactoring * iq1_m: remove unused variable We have progressed to warnings being errors. * iq1_m: add to backend-ops tests * iq1_m: fix Windows ARM * iq1_m: use common definition of iq1m_scale_t * cuda: assert -> NO_DEVICE_CODE * iq1_M: PR comments --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-26 14:21:27 +00:00
case GGML_TYPE_IQ1_M: ftype = LLAMA_FTYPE_MOSTLY_IQ1_M; break;
IQ4_NL: 4-bit non-linear quants with blocks of 32 (#5590) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * iq4_nl: Fix after merging with master * iq4_nl: another fix after merging with master * Use IQ4_NL instead of Q4_K when using k-quants is not possible * Fix typo that makes several tests fail * It was the ggml_vdotq thing missed inside the brackets --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-21 09:39:52 +00:00
case GGML_TYPE_IQ4_NL: ftype = LLAMA_FTYPE_MOSTLY_IQ4_NL; break;
IQ4_XS: a 4.25 bpw quantization (#5747) * Try IQ4_NL with blocks of 64 - does not look good * iq4_xs: go to super-blocks of 256 and 6-bit scales for blocks of 32 * iq4_xs: CUDA works - 133.2 t/s * iq4_xs: AVX2 dot product * iq4_xs: ARM_NEON dot product * iq4_nl: Metal implementation As usual, Metal / Apple Silicon don't like my quants. * iq3_xs: minor fix * iq4_xs: shrink by using IQ3_S for attn_k and attn_q * iq4_xs: revert using IQ3_S for attn_k and attn_v PPL vs size is good, but CPU performance suffers: on M2 Max TG-128 drops to 21.7 t/s from 28.8, and on a Ryzen-7950X to 14.5 t/s from 15.8 t/s. On CUDA we have 135 t/s when using IQ3_S vs 133 t/s with pure IQ4_XS. * Fix CI * iq4_xs: Added forgotten check for 256 divisibility --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-27 14:34:24 +00:00
case GGML_TYPE_IQ4_XS: ftype = LLAMA_FTYPE_MOSTLY_IQ4_XS; break;
IQ3_S: a much better alternative to Q3_K (#5676) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * Resurrecting iq3_xs After all the experimentation, nothing was better than this. * Minor PPL improvement via a block scale fudge factor * Minor improvement via 3 neighbours * iq3_xs: working scalar and AVX2 dot products * iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s) * iq3_xs: working Metal implementation * Adding IQ3_M - IQ3_XS mix with mostly Q4_K * iiq3_xs: a 3.4375 bpw variant * iq3_xs: make CUDA work for new version * iq3_xs: make scalar and AVX2 work for new version * iq3_s: make ARM_NEON work with new version * iq3_xs: make new version work on metal Performance is very similar to Q3_K_S * iq3_xs: tiny Metal speed improvement * iq3_xs: tiny Metal speed improvement * Fix stupid warning * Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS * iq3_xs: rename to iq3_s * iq3_s: make tests pass * Move Q3_K_XS mix to 3.25 bpw * Attempt to fix failing tests * Another attempt to fix the Windows builds * Attempt to fix ROCm * ROCm again * iq3_s: partial fix for QK_K = 64 * iq3_s: make it work on metal for QK_K = 64 Pleasent surprise: the coding was super-block size independent, so all it took was to delete some QK_K == 256 guards. * Will this fix ROCm? --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-24 14:23:52 +00:00
case GGML_TYPE_IQ3_S: ftype = LLAMA_FTYPE_MOSTLY_IQ3_S; break;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
default:
llama : add functions to get the model's metadata (#4013) * llama : add functions to get the model's metadata * format -> std::to_string * better documentation
2023-11-17 15:17:37 +00:00
{
LLAMA_LOG_WARN("%s: unknown type %s\n", __func__, ggml_type_name(type_max));
ftype = LLAMA_FTYPE_ALL_F32;
} break;
llama : fix regression from #2000 - could not load no-mmap models
2023-07-20 10:47:26 +00:00
}
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
gguf : add ftype meta info to the model (#2710) * llama : add ftype meta info to the model ggml-ci * convert.py : add ftype when converting (does not work) * convert.py : fix Enum to IntEnum ggml-ci
2023-08-22 17:05:59 +00:00
// this is a way to mark that we have "guessed" the file type
ftype = (llama_ftype) (ftype | LLAMA_FTYPE_GUESSED);
{
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
const int kid = gguf_find_key(meta, "general.file_type");
gguf : add ftype meta info to the model (#2710) * llama : add ftype meta info to the model ggml-ci * convert.py : add ftype when converting (does not work) * convert.py : fix Enum to IntEnum ggml-ci
2023-08-22 17:05:59 +00:00
if (kid >= 0) {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
ftype = (llama_ftype) gguf_get_val_u32(meta, kid);
gguf : add ftype meta info to the model (#2710) * llama : add ftype meta info to the model ggml-ci * convert.py : add ftype when converting (does not work) * convert.py : fix Enum to IntEnum ggml-ci
2023-08-22 17:05:59 +00:00
}
}
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
LLAMA_LOG_INFO("%s: Dumping metadata keys/values. Note: KV overrides do not apply in this output.\n", __func__);
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
for (int i = 0; i < n_kv; i++) {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
const char * name = gguf_get_key(meta, i);
const enum gguf_type type = gguf_get_kv_type(meta, i);
llama : add functions to get the model's metadata (#4013) * llama : add functions to get the model's metadata * format -> std::to_string * better documentation
2023-11-17 15:17:37 +00:00
const std::string type_name =
type == GGUF_TYPE_ARRAY
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
? format("%s[%s,%d]", gguf_type_name(type), gguf_type_name(gguf_get_arr_type(meta, i)), gguf_get_arr_n(meta, i))
llama : add functions to get the model's metadata (#4013) * llama : add functions to get the model's metadata * format -> std::to_string * better documentation
2023-11-17 15:17:37 +00:00
: gguf_type_name(type);
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
std::string value = gguf_kv_to_str(meta, i);
llama : add functions to get the model's metadata (#4013) * llama : add functions to get the model's metadata * format -> std::to_string * better documentation
2023-11-17 15:17:37 +00:00
const size_t MAX_VALUE_LEN = 40;
if (value.size() > MAX_VALUE_LEN) {
value = format("%s...", value.substr(0, MAX_VALUE_LEN - 3).c_str());
}
gguf-py : export chat templates (#4125) * gguf-py : export chat templates * llama.cpp : escape new lines in gguf kv info prints * gguf-py : bump version * gguf-py : check chat_template type * gguf-py : initialize chat_template
2023-11-19 10:10:52 +00:00
replace_all(value, "\n", "\\n");
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
llama : add functions to get the model's metadata (#4013) * llama : add functions to get the model's metadata * format -> std::to_string * better documentation
2023-11-17 15:17:37 +00:00
LLAMA_LOG_INFO("%s: - kv %3d: %42s %-16s = %s\n", __func__, i, name, type_name.c_str(), value.c_str());
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
// print type counts
for (auto & kv : n_type) {
if (kv.second == 0) {
continue;
}
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
LLAMA_LOG_INFO("%s: - type %4s: %4d tensors\n", __func__, ggml_type_name(kv.first), kv.second);
}
}
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
if (!llama_mmap::SUPPORTED) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
LLAMA_LOG_WARN("%s: mmap is not supported on this platform\n", __func__);
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
use_mmap = false;
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
this->use_mmap = use_mmap;
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
this->check_tensors = check_tensors;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
~llama_model_loader() {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
if (meta) {
gguf_free(meta);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
for (auto * ctx : contexts) {
ggml_free(ctx);
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
}
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
template<typename T>
typename std::enable_if<std::is_integral<T>::value, bool>::type
get_arr_n(const std::string & key, T & result, const bool required = true) {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
const int kid = gguf_find_key(meta, key.c_str());
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
if (kid < 0) {
if (required) {
throw std::runtime_error(format("key not found in model: %s", key.c_str()));
}
return false;
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
struct GGUFMeta::ArrayInfo arr_info =
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
GGUFMeta::GKV<GGUFMeta::ArrayInfo>::get_kv(meta, kid);
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
result = arr_info.length;
return true;
}
template<typename T>
typename std::enable_if<std::is_integral<T>::value, bool>::type
get_arr_n(const enum llm_kv kid, T & result, const bool required = true) {
return get_arr_n(llm_kv(kid), result, required);
}
template<typename T>
bool get_key(const std::string & key, T & result, const bool required = true) {
auto it = kv_overrides.find(key);
const struct llama_model_kv_override * override =
it != kv_overrides.end() ? &it->second : nullptr;
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
const bool found = GGUFMeta::GKV<T>::set(meta, key, result, override);
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
if (required && !found) {
throw std::runtime_error(format("key not found in model: %s", key.c_str()));
}
return found;
}
template<typename T>
bool get_key(const enum llm_kv kid, T & result, const bool required = true) {
return get_key(llm_kv(kid), result, required);
}
std::string get_arch_name() const {
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
return arch_name;
}
enum llm_arch get_arch() const {
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
return llm_kv.arch;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
const char * get_tensor_name(int i) const {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
return weights.at(i).tensor->name;
}
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
const llama_tensor_weight * get_weight(const char * name) const {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
for (const auto & weight : weights) {
if (strcmp(name, weight.tensor->name) == 0) {
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
return &weight;
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
}
}
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
return nullptr;
}
quantize : add '--keep-split' to quantize model into shards (#6688) * Implement '--keep-split' to quantize model into several shards * Add test script * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Split model correctly even if tensor id is out-of-order * Update llama_model_quantize_params * Fix preci failures --------- Co-authored-by: z5269887 <z5269887@unsw.edu.au> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-25 10:29:35 +00:00
const llama_tensor_weight * get_weight(int i) const {
return get_weight(get_tensor_name(i));
}
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
const llama_tensor_weight & require_weight(const char * name) const {
const llama_tensor_weight * weight = get_weight(name);
if (!weight) {
throw std::runtime_error(format("%s: tensor '%s' not found", __func__, name));
}
return *weight;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
Add LoRA support (#820)
2023-04-17 15:28:55 +00:00
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
struct ggml_tensor * get_tensor_meta(const char * name) const {
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
const auto * weight = get_weight(name);
if (!weight) {
return nullptr;
}
return weight->tensor;
}
struct ggml_tensor * require_tensor_meta(const char * name) const {
struct ggml_tensor * tensor = get_tensor_meta(name);
if (!tensor) {
throw std::runtime_error(format("%s: tensor '%s' not found", __func__, name));
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
}
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
return tensor;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
struct ggml_tensor * get_tensor_meta(int i) const {
return get_tensor_meta(get_tensor_name(i));
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
struct ggml_tensor * create_tensor_for(struct ggml_context * ctx, const struct ggml_tensor * cur) {
struct ggml_tensor * tensor = ggml_dup_tensor(ctx, cur);
ggml_set_name(tensor, ggml_get_name(cur));
Multi GPU support, CUDA refactor, CUDA scratch buffer (#1703) * CUDA multi GPU + scratch ggml_cuda_compute_forward Tensor parallelism ggml_cuda_add ggml_cuda_rms_norm ggml_cuda_silu CUDA scratch buffer --main-gpu CLI option
2023-06-06 19:33:23 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
n_created++;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
return tensor;
}
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
const struct ggml_tensor * check_tensor_dims(const std::string & name, const std::vector<int64_t> & ne, bool required) const {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
const struct ggml_tensor * cur = get_tensor_meta(name.c_str());
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
if (cur == NULL) {
llama : avoid using "optional" keyword (#4283)
2023-12-01 18:39:12 +00:00
if (!required) {
llama : support optional tensors (#4283)
2023-12-01 18:35:03 +00:00
return NULL;
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
throw std::runtime_error(format("%s: tensor '%s' not found", __func__, name.c_str()));
}
{
bool is_ok = true;
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
for (size_t i = 0; i < GGML_MAX_DIMS; ++i) {
if ((i < ne.size() && ne[i] != cur->ne[i]) || (i >= ne.size() && cur->ne[i] != 1)) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
is_ok = false;
break;
}
}
if (!is_ok) {
throw std::runtime_error(
format("%s: tensor '%s' has wrong shape; expected %s, got %s",
__func__, name.c_str(),
llama_format_tensor_shape(ne).c_str(),
llama_format_tensor_shape(cur).c_str()));
}
}
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
return cur;
}
struct ggml_tensor * create_tensor(struct ggml_context * ctx, const std::string & name, const std::vector<int64_t> & ne, bool required = true) {
const struct ggml_tensor * cur = check_tensor_dims(name, ne, required);
if (cur == NULL) {
return NULL;
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
return create_tensor_for(ctx, cur);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
struct ggml_tensor * create_tensor_as_view(struct ggml_context * ctx, struct ggml_tensor * base, const std::string & name, const std::vector<int64_t> & ne, size_t offset, bool required = true) {
const struct ggml_tensor * cur = check_tensor_dims(name, ne, required);
if (cur == NULL) {
return NULL;
}
if (cur->type != base->type) {
throw std::runtime_error(format("%s: tensor '%s' has wrong type; expected %s, got %s", __func__, name.c_str(), ggml_type_name(base->type), ggml_type_name(cur->type)));
}
std::array<int64_t, GGML_MAX_DIMS> dims;
for (size_t i = 0; i < GGML_MAX_DIMS; ++i) {
dims[i] = i < ne.size() ? ne[i] : 1;
}
struct ggml_tensor * tensor = ggml_view_4d(ctx, base,
dims[0], dims[1], dims[2], dims[3],
cur->nb[1], cur->nb[2], cur->nb[3],
offset);
ggml_set_name(tensor, name.c_str());
n_created++;
return tensor;
}
llama : fix various warnings
2023-05-13 08:23:15 +00:00
void done_getting_tensors() const {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
if (n_created != n_tensors) {
throw std::runtime_error(format("%s: wrong number of tensors; expected %d, got %d", __func__, n_tensors, n_created));
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
}
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
void init_mappings(bool prefetch = true, llama_mlocks * mlock_mmaps = nullptr) {
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
if (use_mmap) {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
mappings.reserve(files.size());
mmaps_used.reserve(files.size());
for (const auto & file : files) {
std::unique_ptr<llama_mmap> mapping(new llama_mmap(file.get(), prefetch ? -1 : 0, ggml_is_numa()));
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
mmaps_used.emplace_back(mapping->size, 0);
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
if (mlock_mmaps) {
std::unique_ptr<llama_mlock> mlock_mmap(new llama_mlock());
mlock_mmap->init(mapping->addr);
mlock_mmaps->emplace_back(std::move(mlock_mmap));
}
mappings.emplace_back(std::move(mapping));
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
}
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// compute the total size of all tensors for progress reporting
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
for (auto & w : weights) {
size_data += ggml_nbytes(w.tensor);
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
}
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
void get_mapping_range(size_t * first, size_t * last, void ** addr, int idx, ggml_context * ctx) const {
GGML_ASSERT(!mappings.empty());
const auto & mapping = mappings.at(idx);
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
*first = mapping->size;
*last = 0;
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
*addr = mapping->addr;
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (ggml_tensor * tensor = ggml_get_first_tensor(ctx); tensor; tensor = ggml_get_next_tensor(ctx, tensor)) {
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
try {
const auto * weight = get_weight(ggml_get_name(tensor));
if (!weight) {
continue;
}
if (weight->idx != idx) {
continue;
}
*first = std::min(*first, weight->offs);
*last = std::max(*last, weight->offs + ggml_nbytes(tensor));
} catch(...) {
// the tensor is not in the model
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
}
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
}
// for backwards compatibility, does not support ggml-backend
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
void load_data_for(struct ggml_tensor * cur) const {
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
const auto & w = require_weight(ggml_get_name(cur));
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
if (use_mmap) {
const auto & mapping = mappings.at(w.idx);
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
if (cur->data == nullptr) {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
cur->data = (uint8_t *)mapping->addr + w.offs;
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
} else {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
memcpy(cur->data, (uint8_t *)mapping->addr + w.offs, ggml_nbytes(cur));
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
} else {
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
GGML_ASSERT(cur->data != nullptr);
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
GGML_ASSERT(w.idx < files.size());
const auto & file = files.at(w.idx);
file->seek(w.offs, SEEK_SET);
file->read_raw(cur->data, ggml_nbytes(cur));
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
if (check_tensors && !ggml_validate_row_data(cur->type, cur->data, ggml_nbytes(cur))) {
throw std::runtime_error(format("tensor '%s' has invalid data", ggml_get_name(cur)));
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
size_t size_done = 0;
size_t size_data = 0;
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
std::vector<std::pair<size_t, size_t>> mmaps_used;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// Returns false if cancelled by progress_callback
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
bool load_all_data(
struct ggml_context * ctx,
llama_buf_map & bufs_mmap,
llama_mlocks * lmlocks,
llama_progress_callback progress_callback,
void * progress_callback_user_data) {
GGML_ASSERT(size_data != 0 && "call init_mappings() first");
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
std::vector<no_init<uint8_t>> read_buf;
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
std::vector<std::future<std::pair<ggml_tensor *, bool>>> validation_result;
llama : fix loading models with shared tok_embd and output (#5651) ggml-ci
2024-02-21 23:42:09 +00:00
for (struct ggml_tensor * cur = ggml_get_first_tensor(ctx); cur != NULL; cur = ggml_get_next_tensor(ctx, cur)) {
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
const auto * weight = get_weight(ggml_get_name(cur));
if (weight == nullptr) {
// this can happen with split experts models
continue;
}
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
if (progress_callback) {
llama : add ability to cancel model loading (#4462) * llama : Add ability to cancel model load Updated llama_progress_callback so that if it returns false, the model loading is aborted. * llama : Add test for model load cancellation * Fix bool return in llama_model_load, remove std::ignore use * Update llama.cpp Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Fail test if model file is missing * Revert "Fail test if model file is missing" This reverts commit 32ebd525bf7e5a87ee8a3dbaab3d92ce79fbf23d. * Add test-model-load-cancel to Makefile * Revert "Revert "Fail test if model file is missing"" This reverts commit 2796953257ee5383fa7c8fe8fa8fc888c048fb0b. * Simplify .gitignore for tests, clang-tidy fixes * Label all ctest tests * ci : ctest uses -L main * Attempt at writing ctest_with_model * ci : get ci/run.sh working with test-model-load-cancel * ci : restrict .github/workflows/build.yml ctest to -L main * update requirements.txt * Disable test-model-load-cancel in make * Remove venv before creation * Restructure requirements.txt Top-level now imports the specific additional requirements for each python file. Using `pip install -r requirements.txt` will fail if versions become mismatched in the per-file requirements. * Make per-python-script requirements work alone This doesn't break the main requirements.txt. * Add comment * Add convert-persimmon-to-gguf.py to new requirements.txt scheme * Add check-requirements.sh script and GitHub workflow * Remove shellcheck installation step from workflow * Add nocleanup special arg * Fix merge see: https://github.com/ggerganov/llama.cpp/pull/4462#discussion_r1434593573 * reset to upstream/master * Redo changes for cancelling model load --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2023-12-22 06:19:36 +00:00
if (!progress_callback((float) size_done / size_data, progress_callback_user_data)) {
return false;
}
Leverage mmap for offloading tensors to GPU (#1597) * Rebase to latest * Show progress * Add assert to make sure we only allocate temp buffer for non-CPU backend tensor Co-authored-by: Johannes Gäßler <johannesg@5d6.de> --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2023-06-12 12:44:16 +00:00
}
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
size_t n_size = ggml_nbytes(cur);
Leverage mmap for offloading tensors to GPU (#1597) * Rebase to latest * Show progress * Add assert to make sure we only allocate temp buffer for non-CPU backend tensor Co-authored-by: Johannes Gäßler <johannesg@5d6.de> --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2023-06-12 12:44:16 +00:00
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
if (use_mmap) {
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
const auto & mapping = mappings.at(weight->idx);
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
ggml_backend_buffer_t buf_mmap = nullptr;
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
if (bufs_mmap.count(weight->idx)) {
buf_mmap = bufs_mmap.at(weight->idx);
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
}
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
uint8_t * data = (uint8_t *) mapping->addr + weight->offs;
if (check_tensors) {
validation_result.emplace_back(std::async(std::launch::async, [cur, data, n_size] {
return std::make_pair(cur, ggml_validate_row_data(cur->type, data, n_size));
}));
}
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
GGML_ASSERT(buf_mmap || cur->data); // either we have a buffer to allocate the tensor in, or it is already allocated
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
if (buf_mmap && cur->data == nullptr) {
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
ggml_backend_tensor_alloc(buf_mmap, cur, data);
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
if (lmlocks) {
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
const auto & lmlock = lmlocks->at(weight->idx);
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
lmlock->grow_to(weight->offs + n_size);
Leverage mmap for offloading tensors to GPU (#1597) * Rebase to latest * Show progress * Add assert to make sure we only allocate temp buffer for non-CPU backend tensor Co-authored-by: Johannes Gäßler <johannesg@5d6.de> --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2023-06-12 12:44:16 +00:00
}
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
auto & mmap_used = mmaps_used[weight->idx];
mmap_used.first = std::min(mmap_used.first, weight->offs);
mmap_used.second = std::max(mmap_used.second, weight->offs + n_size);
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
} else {
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
ggml_backend_tensor_set(cur, data, 0, n_size);
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
} else {
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
GGML_ASSERT(weight->idx < files.size());
const auto & file = files.at(weight->idx);
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
if (ggml_backend_buffer_is_host(cur->buffer)) {
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
file->seek(weight->offs, SEEK_SET);
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
file->read_raw(cur->data, n_size);
if (check_tensors) {
validation_result.emplace_back(std::async(std::launch::async, [cur, n_size] {
return std::make_pair(cur, ggml_validate_row_data(cur->type, cur->data, n_size));
}));
}
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
} else {
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
read_buf.resize(n_size);
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
file->seek(weight->offs, SEEK_SET);
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
file->read_raw(read_buf.data(), n_size);
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
ggml_backend_tensor_set(cur, read_buf.data(), 0, n_size);
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
if (check_tensors && !ggml_validate_row_data(cur->type, read_buf.data(), n_size)) {
throw std::runtime_error(format("tensor '%s' has invalid data", ggml_get_name(cur)));
}
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
Leverage mmap for offloading tensors to GPU (#1597) * Rebase to latest * Show progress * Add assert to make sure we only allocate temp buffer for non-CPU backend tensor Co-authored-by: Johannes Gäßler <johannesg@5d6.de> --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2023-06-12 12:44:16 +00:00
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
size_done += n_size;
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
// check validation results
bool validation_failed = false;
for (auto & future : validation_result) {
auto result = future.get();
if (!result.second) {
LLAMA_LOG_ERROR("%s: tensor '%s' has invalid data\n", __func__, ggml_get_name(result.first));
validation_failed = true;
}
}
if (validation_failed) {
throw std::runtime_error("found tensors with invalid data");
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// check if this is the last call and do final cleanup
if (size_done >= size_data) {
// unmap offloaded tensors and metadata
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
if (use_mmap) {
for (uint32_t idx = 0; idx < mappings.size(); idx++) {
const auto & mmap_used = mmaps_used.at(idx);
auto & mapping = mappings.at(idx);
mapping->unmap_fragment(0, mmap_used.first);
if (mmap_used.second != 0) {
mapping->unmap_fragment(mmap_used.second, mapping->size);
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
}
}
if (progress_callback) {
// Even though the model is done loading, we still honor
// cancellation since we need to free allocations.
return progress_callback(1.0f, progress_callback_user_data);
}
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
llama : add ability to cancel model loading (#4462) * llama : Add ability to cancel model load Updated llama_progress_callback so that if it returns false, the model loading is aborted. * llama : Add test for model load cancellation * Fix bool return in llama_model_load, remove std::ignore use * Update llama.cpp Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Fail test if model file is missing * Revert "Fail test if model file is missing" This reverts commit 32ebd525bf7e5a87ee8a3dbaab3d92ce79fbf23d. * Add test-model-load-cancel to Makefile * Revert "Revert "Fail test if model file is missing"" This reverts commit 2796953257ee5383fa7c8fe8fa8fc888c048fb0b. * Simplify .gitignore for tests, clang-tidy fixes * Label all ctest tests * ci : ctest uses -L main * Attempt at writing ctest_with_model * ci : get ci/run.sh working with test-model-load-cancel * ci : restrict .github/workflows/build.yml ctest to -L main * update requirements.txt * Disable test-model-load-cancel in make * Remove venv before creation * Restructure requirements.txt Top-level now imports the specific additional requirements for each python file. Using `pip install -r requirements.txt` will fail if versions become mismatched in the per-file requirements. * Make per-python-script requirements work alone This doesn't break the main requirements.txt. * Add comment * Add convert-persimmon-to-gguf.py to new requirements.txt scheme * Add check-requirements.sh script and GitHub workflow * Remove shellcheck installation step from workflow * Add nocleanup special arg * Fix merge see: https://github.com/ggerganov/llama.cpp/pull/4462#discussion_r1434593573 * reset to upstream/master * Redo changes for cancelling model load --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2023-12-22 06:19:36 +00:00
return true;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
};
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
template<>
bool llama_model_loader::get_key(const enum llm_kv kid, enum llama_pooling_type & result, const bool required) {
uint32_t tmp;
const bool found = get_key(kid, tmp, required);
llama : allow for user specified embedding pooling type (#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-03 10:40:27 +00:00
if (found) {
result = (enum llama_pooling_type) tmp;
} else {
result = LLAMA_POOLING_TYPE_UNSPECIFIED;
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
return found;
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
//
// load LLaMA models
//
llama : remove shards weight file support (#2000) * Remove multiple shards * Remove multiple file loaders * Remove llama_load_tensor_shard class * Simplify load logic * Remove dead code guess_n_parts function * Remove vocab_only from constructor of llama_model_loader * Remove alignment_prevents_mmap which is not more needed. * Remove useless check
2023-06-28 17:13:02 +00:00
YaRN : store rope scaling type as int32_t in memory (#5285) * YaRN : store rope scaling type as int32_t in memory * llama : store mapped names as const char *
2024-02-03 11:22:06 +00:00
static const char * llama_model_arch_name(llm_arch arch) {
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
auto it = LLM_ARCH_NAMES.find(arch);
if (it == LLM_ARCH_NAMES.end()) {
return "unknown";
}
return it->second;
}
static std::string llama_model_ftype_name(llama_ftype ftype) {
gguf : add ftype meta info to the model (#2710) * llama : add ftype meta info to the model ggml-ci * convert.py : add ftype when converting (does not work) * convert.py : fix Enum to IntEnum ggml-ci
2023-08-22 17:05:59 +00:00
if (ftype & LLAMA_FTYPE_GUESSED) {
return llama_model_ftype_name((enum llama_ftype) (ftype & ~LLAMA_FTYPE_GUESSED)) + " (guessed)";
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
switch (ftype) {
case LLAMA_FTYPE_ALL_F32: return "all F32";
llama.swiftui : add bench functionality (#4483) * llama.swiftui : add bench button * llama.swiftui : initial bench functionality * force to use n_gpu_layers on simulator * add download buttons & expose llamaState.loadModel * update project.pbxproj * comment #Preview & fix editorconfig check * gitignore : xcode stuff * llama.swiftui : UX improvements * llama.swiftui : avoid data copy via "downloadTask" * llama.swiftui : remove model from project * llama : remove "mostly" from model infos * llama.swiftui : improve bench --------- Co-authored-by: jhen <developer@jhen.me>
2023-12-17 17:38:41 +00:00
case LLAMA_FTYPE_MOSTLY_F16: return "F16";
case LLAMA_FTYPE_MOSTLY_Q4_0: return "Q4_0";
case LLAMA_FTYPE_MOSTLY_Q4_1: return "Q4_1";
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
case LLAMA_FTYPE_MOSTLY_Q4_1_SOME_F16:
llama.swiftui : add bench functionality (#4483) * llama.swiftui : add bench button * llama.swiftui : initial bench functionality * force to use n_gpu_layers on simulator * add download buttons & expose llamaState.loadModel * update project.pbxproj * comment #Preview & fix editorconfig check * gitignore : xcode stuff * llama.swiftui : UX improvements * llama.swiftui : avoid data copy via "downloadTask" * llama.swiftui : remove model from project * llama : remove "mostly" from model infos * llama.swiftui : improve bench --------- Co-authored-by: jhen <developer@jhen.me>
2023-12-17 17:38:41 +00:00
return "Q4_1, some F16";
case LLAMA_FTYPE_MOSTLY_Q5_0: return "Q5_0";
case LLAMA_FTYPE_MOSTLY_Q5_1: return "Q5_1";
case LLAMA_FTYPE_MOSTLY_Q8_0: return "Q8_0";
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
// K-quants
llama : restore intended k-quants mixes for MoE models (#4872) * Restore intended k-quants quantization mixes for MoE models * Update Q2_K_S values in the quantize tool Still using LLaMA-v1 PPL values in the quant description today does not make much sense. But let's leave this update for another PR. --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-11 19:43:15 +00:00
case LLAMA_FTYPE_MOSTLY_Q2_K: return "Q2_K - Medium";
case LLAMA_FTYPE_MOSTLY_Q2_K_S: return "Q2_K - Small";
llama.swiftui : add bench functionality (#4483) * llama.swiftui : add bench button * llama.swiftui : initial bench functionality * force to use n_gpu_layers on simulator * add download buttons & expose llamaState.loadModel * update project.pbxproj * comment #Preview & fix editorconfig check * gitignore : xcode stuff * llama.swiftui : UX improvements * llama.swiftui : avoid data copy via "downloadTask" * llama.swiftui : remove model from project * llama : remove "mostly" from model infos * llama.swiftui : improve bench --------- Co-authored-by: jhen <developer@jhen.me>
2023-12-17 17:38:41 +00:00
case LLAMA_FTYPE_MOSTLY_Q3_K_S: return "Q3_K - Small";
case LLAMA_FTYPE_MOSTLY_Q3_K_M: return "Q3_K - Medium";
case LLAMA_FTYPE_MOSTLY_Q3_K_L: return "Q3_K - Large";
case LLAMA_FTYPE_MOSTLY_Q4_K_S: return "Q4_K - Small";
case LLAMA_FTYPE_MOSTLY_Q4_K_M: return "Q4_K - Medium";
case LLAMA_FTYPE_MOSTLY_Q5_K_S: return "Q5_K - Small";
case LLAMA_FTYPE_MOSTLY_Q5_K_M: return "Q5_K - Medium";
case LLAMA_FTYPE_MOSTLY_Q6_K: return "Q6_K";
Fix typos of IQ2_XXS and IQ3_XXS in llama.cpp (#5231)
2024-01-31 03:04:21 +00:00
case LLAMA_FTYPE_MOSTLY_IQ2_XXS:return "IQ2_XXS - 2.0625 bpw";
ggml : SOTA 2-bit quants (add IQ2_XS) (#4856) * iq2_xs: basics * iq2_xs: this should have been in the basics * iq2_xs: CUDA and scalar CPU works * iq2_xs: WIP Metal * iq2_xs: Metal now works * iq2_xs: working, but dog slow, ARM_NEON dot product * iq2_xs: better ARM_NEON dot product We are now at 19.5 t/s for TG-128 and 61 t/s for PP-512 when running on the CPU. * iq2_xs: AVX2 dot product - 19.5 t/s * iq2_xs: faster AVX2 dit product 21.4 t/s for TG-128, 59.2 t/s for PP-512. The latter is 2x compared to the previous version. * iq2_xs: had forgotten to delete iq2-data.h * Add llama enum for IQ2_XS --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-11 19:39:39 +00:00
case LLAMA_FTYPE_MOSTLY_IQ2_XS: return "IQ2_XS - 2.3125 bpw";
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
case LLAMA_FTYPE_MOSTLY_IQ2_S: return "IQ2_S - 2.5 bpw";
case LLAMA_FTYPE_MOSTLY_IQ2_M: return "IQ2_M - 2.7 bpw";
case LLAMA_FTYPE_MOSTLY_IQ3_XS: return "IQ3_XS - 3.3 bpw";
Fix typos of IQ2_XXS and IQ3_XXS in llama.cpp (#5231)
2024-01-31 03:04:21 +00:00
case LLAMA_FTYPE_MOSTLY_IQ3_XXS:return "IQ3_XXS - 3.0625 bpw";
1.5 bit quantization (#5453) * iq1_s: WIP basics * iq1_s: CUDA is working * iq1_s: scalar CPU dot product * iq1_s: WIP AVX2 dot product - something is not right * Fix tests * Fix shadow warnings * Fix after merge with latest master * iq1_s: AVX2 finally works * iq1_s: ARM_NEON dot product. Works, but not very fast * iq1_s: better grid * iq1_s: use IQ2_XXS for attn_output At a cost of 0.04 extra bpw this gives a big improvement in PPL. * iq1_s: Metal basics Dequantize works, but not dot product * iq1_s: Metal works, but quite slow As usual, Apple Silicon does not like the code I write. * iq1_s: Tests * iq1_s: slightly faster dot product --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-18 16:16:55 +00:00
case LLAMA_FTYPE_MOSTLY_IQ1_S :return "IQ1_S - 1.5625 bpw";
IQ1_M: 1.75 bpw quantization (#6302) * iq1_m: basics * iq1_m: basics-2 * iq1_m: CUDA dequantize works Very 1st shot I get PPL = 9.76 for LLaMA-v2-7B. * iq1_m: separate shifts for each group of 8 in a block We get PPL(LLaMA-v2-7B ) = 9.2810 PPL(LLaMA-v2-13B) = 6.8105 Not bad, but slightly higher than sqrt(PPL(IQ1_S) * PPL(IQ2_XXS)) which is the expected outcome given that IQ1_M is halfway between IQ1_S and IQ2_XXS in terms of bpw. From this, we would expect PPL = 9.14 for LLaMA-v2-7B PPL = 6.63 for LLaMA-v2-13B * iq1_m: go to 3-bit scales There is slight increase in PPL, but the 0.0625 bpw reduction in size is totally worth it. We now have PPL(LLaMA-v2-7B ) = 9.4469 at 1.96 bpw PPL(LLaMA-v2-13B) = 6.8717 at 1.93 bpw PPL(LLaMA-v2-70B) = 4.8568 at 1.85 bpw * iq1_m: scalar dot product * iq1_m: AVX2 dot product * iq1_m: very slightly faster AVX2 dot product * iq1_m: ARM_NEON dot product Works, but very slow (10.5 t/s) * iq1_m: Metal - dequantize works, dot product does not * iq1_m: Metal now works About the same performance as iq1_s. * iq1_m: minor * iq1_m: checking pure iq1_m quantization It is pretty bad: PPL(LLaMA-v2-7B) = 34 if we quantize output.weight with Q4_K. * iiq1_m: slightly faster ARM_NEON dot product 10.5 t/s -> 11.65 t/s * iq1_m: faster ARM_NEON dot product 11.65 t/s -> 14.9 t/s * iq1_m: another minor ARM_NEON dot product improvement 14.9 -> 15.0 t/s * iq1_m: small PPL improvement via super-block scale adjustment After quantizing block scales redo the super-block scale fit. PPL(LLaMA-v2-7B ) = 9.3346 PPL(LLaMA-v2-13B) = 6.8419 PPL(LLaMA-v2-70B) = 4.8294 PPL(Mistral-7B ) = 8.1624 * iq1_m: adapt to CUDA refactoring * iq1_m: remove unused variable We have progressed to warnings being errors. * iq1_m: add to backend-ops tests * iq1_m: fix Windows ARM * iq1_m: use common definition of iq1m_scale_t * cuda: assert -> NO_DEVICE_CODE * iq1_M: PR comments --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-26 14:21:27 +00:00
case LLAMA_FTYPE_MOSTLY_IQ1_M :return "IQ1_M - 1.75 bpw";
IQ4_NL: 4-bit non-linear quants with blocks of 32 (#5590) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * iq4_nl: Fix after merging with master * iq4_nl: another fix after merging with master * Use IQ4_NL instead of Q4_K when using k-quants is not possible * Fix typo that makes several tests fail * It was the ggml_vdotq thing missed inside the brackets --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-21 09:39:52 +00:00
case LLAMA_FTYPE_MOSTLY_IQ4_NL: return "IQ4_NL - 4.5 bpw";
IQ4_XS: a 4.25 bpw quantization (#5747) * Try IQ4_NL with blocks of 64 - does not look good * iq4_xs: go to super-blocks of 256 and 6-bit scales for blocks of 32 * iq4_xs: CUDA works - 133.2 t/s * iq4_xs: AVX2 dot product * iq4_xs: ARM_NEON dot product * iq4_nl: Metal implementation As usual, Metal / Apple Silicon don't like my quants. * iq3_xs: minor fix * iq4_xs: shrink by using IQ3_S for attn_k and attn_q * iq4_xs: revert using IQ3_S for attn_k and attn_v PPL vs size is good, but CPU performance suffers: on M2 Max TG-128 drops to 21.7 t/s from 28.8, and on a Ryzen-7950X to 14.5 t/s from 15.8 t/s. On CUDA we have 135 t/s when using IQ3_S vs 133 t/s with pure IQ4_XS. * Fix CI * iq4_xs: Added forgotten check for 256 divisibility --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-27 14:34:24 +00:00
case LLAMA_FTYPE_MOSTLY_IQ4_XS: return "IQ4_XS - 4.25 bpw";
IQ3_S: a much better alternative to Q3_K (#5676) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * Resurrecting iq3_xs After all the experimentation, nothing was better than this. * Minor PPL improvement via a block scale fudge factor * Minor improvement via 3 neighbours * iq3_xs: working scalar and AVX2 dot products * iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s) * iq3_xs: working Metal implementation * Adding IQ3_M - IQ3_XS mix with mostly Q4_K * iiq3_xs: a 3.4375 bpw variant * iq3_xs: make CUDA work for new version * iq3_xs: make scalar and AVX2 work for new version * iq3_s: make ARM_NEON work with new version * iq3_xs: make new version work on metal Performance is very similar to Q3_K_S * iq3_xs: tiny Metal speed improvement * iq3_xs: tiny Metal speed improvement * Fix stupid warning * Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS * iq3_xs: rename to iq3_s * iq3_s: make tests pass * Move Q3_K_XS mix to 3.25 bpw * Attempt to fix failing tests * Another attempt to fix the Windows builds * Attempt to fix ROCm * ROCm again * iq3_s: partial fix for QK_K = 64 * iq3_s: make it work on metal for QK_K = 64 Pleasent surprise: the coding was super-block size independent, so all it took was to delete some QK_K == 256 guards. * Will this fix ROCm? --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-24 14:23:52 +00:00
case LLAMA_FTYPE_MOSTLY_IQ3_S: return "IQ3_S - 3.4375 bpw";
case LLAMA_FTYPE_MOSTLY_IQ3_M: return "IQ3_S mix - 3.66 bpw";
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
default: return "unknown, may not work";
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
static const char * llama_model_type_name(e_model type) {
switch (type) {
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
case MODEL_22M: return "22M";
case MODEL_33M: return "33M";
case MODEL_109M: return "109M";
case MODEL_137M: return "137M";
case MODEL_0_5B: return "0.5B";
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
case MODEL_1B: return "1B";
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
case MODEL_2B: return "2B";
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
case MODEL_3B: return "3B";
case MODEL_7B: return "7B";
case MODEL_8B: return "8B";
llama : add StableLM2 12B (#6635) * StableLM2 12B support for huggingface -> GGUF * StableLM12 tensormapping and constants * StableLM-2-12b model support * fix * Added 12B support * Removed autoformatting; resolved bug where model_arch was not selecting StableLM2 * Formatting * Do QK norm stacking in model conversion step * Converge StableLM and StableLM2 code to simplify graph construction * Fix accidental removal * Removed warnings * Revert formatter * Move QK norm stack to private function so it's easier to read * refactor stablelm graph builder to support 1.6, 3b and 12b more efficiently * Proper check for None type for new_name to avoid crash; formatting; revert change to base class `write_tensors()` * Format * Formatting * format Co-authored-by: compilade <git@compilade.net> * Fix incorrect check for K norm * space after commas; Keep indentation multiple of 4 spaces * Flake8 format * Removed unnecessary conditional branches * Removed unused comment * Fixed incorrect tensor passing * Format --------- Co-authored-by: compilade <git@compilade.net>
2024-04-16 15:48:35 +00:00
case MODEL_12B: return "12B";
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
case MODEL_13B: return "13B";
llama : add support for Orion-14B (#5118) * add support for Orion-14B(https://huggingface.co/OrionStarAI/Orion-14B-Chat) * flake8 support * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * Update llama.cpp * Update llama.cpp --------- Co-authored-by: lixiaopu <lixiaopu@cmcm.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-01-28 08:00:30 +00:00
case MODEL_14B: return "14B";
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
case MODEL_15B: return "15B";
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
case MODEL_20B: return "20B";
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
case MODEL_30B: return "30B";
case MODEL_34B: return "34B";
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
case MODEL_35B: return "35B";
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
case MODEL_40B: return "40B";
case MODEL_65B: return "65B";
case MODEL_70B: return "70B";
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
case MODEL_314B: return "314B";
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
case MODEL_SMALL: return "0.1B";
case MODEL_MEDIUM: return "0.4B";
case MODEL_LARGE: return "0.8B";
case MODEL_XL: return "1.5B";
llama : add qwen2moe (#6074) * support qwen2moe * fix-review * metal : support unary ops for nelements % 4 != 0 * metal : require contiguousness for float4 unary kernels * metal : require contiguousness for float4 unary kernels (cont) * fix-review * names : for brevity "SHARED_EXP" -> "SHEXP" * llama : reuse build_moe_ffn() * llama : add model type name --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-16 15:40:48 +00:00
case MODEL_A2_7B: return "A2.7B";
llama : add model types for mixtral (#6589)
2024-04-10 15:24:14 +00:00
case MODEL_8x7B: return "8x7B";
case MODEL_8x22B: return "8x22B";
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
case MODEL_16x12B: return "16x12B";
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
default: return "?B";
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
static const char * llama_model_vocab_type_name(enum llama_vocab_type type){
switch (type) {
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
case LLAMA_VOCAB_TYPE_NONE: return "no vocab";
case LLAMA_VOCAB_TYPE_SPM: return "SPM";
case LLAMA_VOCAB_TYPE_BPE: return "BPE";
case LLAMA_VOCAB_TYPE_WPM: return "WPM";
default: return "unknown";
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
}
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
static void llm_load_arch(llama_model_loader & ml, llama_model & model) {
model.arch = ml.get_arch();
if (model.arch == LLM_ARCH_UNKNOWN) {
throw std::runtime_error("unknown model architecture: '" + ml.get_arch_name() + "'");
}
}
static void llm_load_hparams(
llama_model_loader & ml,
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
llama_model & model) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
auto & hparams = model.hparams;
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
const gguf_context * ctx = ml.meta;
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
llama : add functions to get the model's metadata (#4013) * llama : add functions to get the model's metadata * format -> std::to_string * better documentation
2023-11-17 15:17:37 +00:00
// get metadata as string
for (int i = 0; i < gguf_get_n_kv(ctx); i++) {
enum gguf_type type = gguf_get_kv_type(ctx, i);
if (type == GGUF_TYPE_ARRAY) {
continue;
}
const char * name = gguf_get_key(ctx, i);
const std::string value = gguf_kv_to_str(ctx, i);
model.gguf_kv.emplace(name, value);
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// get general kv
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
ml.get_key(LLM_KV_GENERAL_NAME, model.name, false);
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// get hparams kv
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
ml.get_key(LLM_KV_VOCAB_SIZE, hparams.n_vocab, false) || ml.get_arr_n(LLM_KV_TOKENIZER_LIST, hparams.n_vocab);
ml.get_key(LLM_KV_CONTEXT_LENGTH, hparams.n_ctx_train);
ml.get_key(LLM_KV_EMBEDDING_LENGTH, hparams.n_embd);
ml.get_key(LLM_KV_FEED_FORWARD_LENGTH, hparams.n_ff);
ml.get_key(LLM_KV_ATTENTION_HEAD_COUNT, hparams.n_head);
ml.get_key(LLM_KV_BLOCK_COUNT, hparams.n_layer);
ml.get_key(LLM_KV_EXPERT_COUNT, hparams.n_expert, false);
ml.get_key(LLM_KV_EXPERT_USED_COUNT, hparams.n_expert_used, false);
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
GGML_ASSERT(hparams.n_expert <= LLAMA_MAX_EXPERTS);
GGML_ASSERT(hparams.n_expert_used <= hparams.n_expert);
if (hparams.n_expert > 0) {
GGML_ASSERT(hparams.n_expert_used > 0);
} else {
GGML_ASSERT(hparams.n_expert_used == 0);
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// n_head_kv is optional, default to n_head
hparams.n_head_kv = hparams.n_head;
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
ml.get_key(LLM_KV_ATTENTION_HEAD_COUNT_KV, hparams.n_head_kv, false);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
bool rope_finetuned = false;
ml.get_key(LLM_KV_ROPE_SCALING_FINETUNED, rope_finetuned, false);
hparams.rope_finetuned = rope_finetuned;
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
hparams.n_yarn_orig_ctx = hparams.n_ctx_train;
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
ml.get_key(LLM_KV_ROPE_SCALING_ORIG_CTX_LEN, hparams.n_yarn_orig_ctx, false);
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
llama : allow gguf RoPE keys to be overridden with defaults (#3240)
2023-09-20 16:12:47 +00:00
// rope_freq_base (optional)
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
hparams.rope_freq_base_train = 10000.0f;
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
ml.get_key(LLM_KV_ROPE_FREQ_BASE, hparams.rope_freq_base_train, false);
gguf : add rope_freq_base parameter for CodeLlama (#2769)
2023-08-24 18:04:05 +00:00
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
std::string rope_scaling("linear");
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
ml.get_key(LLM_KV_ROPE_SCALING_TYPE, rope_scaling, false);
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
hparams.rope_scaling_type_train = llama_rope_scaling_type_from_string(rope_scaling);
code : normalize enum names (#5697) * coda : normalize enum names ggml-ci * code : cont * code : cont
2024-02-25 10:09:09 +00:00
GGML_ASSERT(hparams.rope_scaling_type_train != LLAMA_ROPE_SCALING_TYPE_UNSPECIFIED);
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// rope_freq_scale (inverse of the kv) is optional
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
float ropescale = 0.0f;
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
if (!ml.get_key(LLM_KV_ROPE_SCALING_FACTOR, ropescale, false)) {
// try the old key name
ml.get_key(LLM_KV_ROPE_SCALE_LINEAR, ropescale, false);
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
}
hparams.rope_freq_scale_train = ropescale == 0.0f ? 1.0f : 1.0f/ropescale;
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// sanity check for n_rot (optional)
{
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
hparams.n_rot = (hparams.n_head == 0) ? 0 : hparams.n_embd / hparams.n_head;
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
ml.get_key(LLM_KV_ROPE_DIMENSION_COUNT, hparams.n_rot, false);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
examples : fix gpt-neox (#2943) Co-authored-by: mmnga <mmnga1mmnga@gmail.com>
2023-09-03 05:36:28 +00:00
if (model.arch == LLM_ARCH_LLAMA || model.arch == LLM_ARCH_FALCON) {
if (hparams.n_rot != hparams.n_embd / hparams.n_head) {
throw std::runtime_error(format("invalid n_rot: %u, expected %u", hparams.n_rot, hparams.n_embd / hparams.n_head));
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
examples : fix gpt-neox (#2943) Co-authored-by: mmnga <mmnga1mmnga@gmail.com>
2023-09-03 05:36:28 +00:00
// gpt-neox n_rot = rotary_pct * (n_embd / n_head)
// gpt-j n_rot = rotary_dim
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
hparams.n_embd_head_k = (hparams.n_head == 0) ? 0 : hparams.n_embd / hparams.n_head;
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
ml.get_key(LLM_KV_ATTENTION_KEY_LENGTH, hparams.n_embd_head_k, false);
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
hparams.n_embd_head_v = (hparams.n_head == 0) ? 0 : hparams.n_embd / hparams.n_head;
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
ml.get_key(LLM_KV_ATTENTION_VALUE_LENGTH, hparams.n_embd_head_v, false);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// arch-specific KVs
switch (model.arch) {
case LLM_ARCH_LLAMA:
{
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : add model types for mixtral (#6589)
2024-04-10 15:24:14 +00:00
if (hparams.n_expert == 8) {
switch (hparams.n_layer) {
case 32: model.type = e_model::MODEL_8x7B; break;
case 56: model.type = e_model::MODEL_8x22B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} else {
switch (hparams.n_layer) {
case 22: model.type = e_model::MODEL_1B; break;
case 26: model.type = e_model::MODEL_3B; break;
llama : support Llama 3 HF conversion (#6745) * Support Llama 3 conversion The tokenizer is BPE. * style * Accept suggestion Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> * llama : add llama_token_is_eog() ggml-ci * llama : auto-detect more EOT tokens when missing in KV data * convert : replacing EOS token is a hack * llama : fix codegemma EOT token + add TODOs * llama : fix model type string for 8B model --------- Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-21 11:50:41 +00:00
case 32: model.type = hparams.n_head == hparams.n_head_kv ? e_model::MODEL_7B : e_model::MODEL_8B; break; // LLaMa 8B v3 uses GQA
llama : add model types for mixtral (#6589)
2024-04-10 15:24:14 +00:00
case 40: model.type = e_model::MODEL_13B; break;
case 48: model.type = e_model::MODEL_34B; break;
case 60: model.type = e_model::MODEL_30B; break;
case 80: model.type = hparams.n_head == hparams.n_head_kv ? e_model::MODEL_65B : e_model::MODEL_70B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
} break;
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
case LLM_ARCH_MINICPM:
{
llama : fix MiniCPM (#5392) * fix bug for norm_rms_eps missing * to align with the same order as convert.py for model write * fix: undo HF models permute tensor * update for flake8 lint
2024-02-08 10:36:19 +00:00
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
switch (hparams.n_layer) {
case 40: model.type = e_model::MODEL_2B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
case LLM_ARCH_GROK:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
switch (hparams.n_layer) {
case 64: model.type = e_model::MODEL_314B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
case LLM_ARCH_FALCON:
{
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
switch (hparams.n_layer) {
case 32: model.type = e_model::MODEL_7B; break;
case 60: model.type = e_model::MODEL_40B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
case LLM_ARCH_BAICHUAN:
{
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
switch (hparams.n_layer) {
case 32: model.type = e_model::MODEL_7B; break;
case 40: model.type = e_model::MODEL_13B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
if (model.type == e_model::MODEL_13B) {
// TODO: become GGUF KV parameter
hparams.f_max_alibi_bias = 8.0f;
}
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
} break;
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
case LLM_ARCH_STARCODER:
{
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
switch (hparams.n_layer) {
case 24: model.type = e_model::MODEL_1B; break;
case 36: model.type = e_model::MODEL_3B; break;
case 42: model.type = e_model::MODEL_7B; break;
case 40: model.type = e_model::MODEL_15B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
case LLM_ARCH_PERSIMMON:
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
{
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
switch (hparams.n_layer) {
case 36: model.type = e_model::MODEL_8B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
case LLM_ARCH_REFACT:
{
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
switch (hparams.n_layer) {
case 32: model.type = e_model::MODEL_1B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
// TODO: become GGUF KV parameter
hparams.f_max_alibi_bias = 8.0f;
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
} break;
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
case LLM_ARCH_BERT:
{
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
ml.get_key(LLM_KV_ATTENTION_CAUSAL, hparams.causal_attn);
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
ml.get_key(LLM_KV_TOKENIZER_TOKEN_TYPE_COUNT, hparams.n_vocab_type);
llama : allow for user specified embedding pooling type (#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-03 10:40:27 +00:00
ml.get_key(LLM_KV_POOLING_TYPE, hparams.pooling_type, false);
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
switch (hparams.n_layer) {
case 3:
model.type = e_model::MODEL_17M; break; // bge-micro
case 6:
model.type = e_model::MODEL_22M; break; // MiniLM-L6
case 12:
switch (hparams.n_embd) {
case 384: model.type = e_model::MODEL_33M; break; // MiniLM-L12, bge-small
case 768: model.type = e_model::MODEL_109M; break; // bge-base
} break;
case 24:
model.type = e_model::MODEL_335M; break; // bge-large
}
} break;
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
case LLM_ARCH_NOMIC_BERT:
{
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
ml.get_key(LLM_KV_ATTENTION_CAUSAL, hparams.causal_attn);
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
ml.get_key(LLM_KV_TOKENIZER_TOKEN_TYPE_COUNT, hparams.n_vocab_type);
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ml.get_key(LLM_KV_POOLING_TYPE, hparams.pooling_type);
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
if (hparams.n_layer == 12 && hparams.n_embd == 768) {
model.type = e_model::MODEL_137M;
}
} break;
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
case LLM_ARCH_BLOOM:
{
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
switch (hparams.n_layer) {
case 24: model.type = e_model::MODEL_1B; break;
case 30:
switch (hparams.n_embd) {
case 2560: model.type = e_model::MODEL_3B; break;
case 4096: model.type = e_model::MODEL_7B; break;
} break;
}
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
// TODO: become GGUF KV parameter
hparams.f_max_alibi_bias = 8.0f;
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
} break;
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
case LLM_ARCH_MPT:
{
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
ml.get_key(LLM_KV_ATTENTION_CLAMP_KQV, hparams.f_clamp_kqv, false);
ml.get_key(LLM_KV_ATTENTION_MAX_ALIBI_BIAS, hparams.f_max_alibi_bias);
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
switch (hparams.n_layer) {
case 32: model.type = e_model::MODEL_7B; break;
case 48: model.type = e_model::MODEL_30B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
case LLM_ARCH_STABLELM:
{
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
switch (hparams.n_layer) {
llama : support StableLM 2 1.6B (#5052) * llama : support StableLM 2 1.6B * convert : fix Qwen's set_vocab wrongly naming all special tokens [PAD{id}] * convert : refactor Qwen's set_vocab to use it for StableLM 2 too * nix : add tiktoken to llama-python-extra * convert : use presence of tokenizer.json to determine StableLM tokenizer loader It's a less arbitrary heuristic than the vocab size.
2024-01-22 11:21:52 +00:00
case 24: model.type = e_model::MODEL_1B; break;
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
case 32: model.type = e_model::MODEL_3B; break;
llama : add StableLM2 12B (#6635) * StableLM2 12B support for huggingface -> GGUF * StableLM12 tensormapping and constants * StableLM-2-12b model support * fix * Added 12B support * Removed autoformatting; resolved bug where model_arch was not selecting StableLM2 * Formatting * Do QK norm stacking in model conversion step * Converge StableLM and StableLM2 code to simplify graph construction * Fix accidental removal * Removed warnings * Revert formatter * Move QK norm stack to private function so it's easier to read * refactor stablelm graph builder to support 1.6, 3b and 12b more efficiently * Proper check for None type for new_name to avoid crash; formatting; revert change to base class `write_tensors()` * Format * Formatting * format Co-authored-by: compilade <git@compilade.net> * Fix incorrect check for K norm * space after commas; Keep indentation multiple of 4 spaces * Flake8 format * Removed unnecessary conditional branches * Removed unused comment * Fixed incorrect tensor passing * Format --------- Co-authored-by: compilade <git@compilade.net>
2024-04-16 15:48:35 +00:00
case 40: model.type = e_model::MODEL_12B; break;
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
case LLM_ARCH_QWEN:
{
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
switch (hparams.n_layer) {
case 32: model.type = e_model::MODEL_7B; break;
case 40: model.type = e_model::MODEL_13B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
llama : support upcoming Qwen2 (#5037)
2024-01-19 11:53:13 +00:00
case LLM_ARCH_QWEN2:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
switch (hparams.n_layer) {
llama : add more qwen2 models (#5071)
2024-01-22 07:33:19 +00:00
case 24: model.type = hparams.n_embd == 1024 ? e_model::MODEL_0_5B : e_model::MODEL_1B; break;
llama : support upcoming Qwen2 (#5037)
2024-01-19 11:53:13 +00:00
case 32: model.type = e_model::MODEL_7B; break;
llama : add more qwen2 models (#5071)
2024-01-22 07:33:19 +00:00
case 40: model.type = hparams.n_head == 20 ? e_model::MODEL_4B : e_model::MODEL_13B; break;
llama : support upcoming Qwen2 (#5037)
2024-01-19 11:53:13 +00:00
case 80: model.type = e_model::MODEL_70B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
llama : add qwen2moe (#6074) * support qwen2moe * fix-review * metal : support unary ops for nelements % 4 != 0 * metal : require contiguousness for float4 unary kernels * metal : require contiguousness for float4 unary kernels (cont) * fix-review * names : for brevity "SHARED_EXP" -> "SHEXP" * llama : reuse build_moe_ffn() * llama : add model type name --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-16 15:40:48 +00:00
case LLM_ARCH_QWEN2MOE:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
switch (hparams.n_layer) {
case 24: model.type = e_model::MODEL_A2_7B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
case LLM_ARCH_PHI2:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
llama : add phi3 support (#6852) * add explicit phi3 support * add explicit phi3 support * remove unused code * convert : add BOS token * llama : match EOT token <|end|> * llama : minor / style * llama : tabs -> spaces * convert : fix lint checks --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-24 07:00:37 +00:00
switch (hparams.n_layer) {
case 24: model.type = e_model::MODEL_1B; break;
case 32: model.type = e_model::MODEL_3B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
case LLM_ARCH_PHI3:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
switch (hparams.n_layer) {
llama : recognize 1B phi models (#4847) This update categorizes models with 24 layers as MODEL_1B, ensuring compatibility with different Phi model variants without impacting existing Phi-2 model functionality.
2024-01-10 13:39:09 +00:00
case 24: model.type = e_model::MODEL_1B; break;
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
case 32: model.type = e_model::MODEL_3B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
case LLM_ARCH_PLAMO:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
switch (hparams.n_layer) {
case 40: model.type = e_model::MODEL_13B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
case LLM_ARCH_GPT2:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
switch (hparams.n_layer) {
case 12: model.type = e_model::MODEL_SMALL; break;
case 24: model.type = e_model::MODEL_MEDIUM; break;
case 36: model.type = e_model::MODEL_LARGE; break;
case 48: model.type = e_model::MODEL_XL; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
llama : add CodeShell support (#5016) * llama: add codeshell support * llama.cpp: fix codeshell with NeoX rope Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-19 09:07:27 +00:00
case LLM_ARCH_CODESHELL:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
switch (hparams.n_layer) {
case 42: model.type = e_model::MODEL_SMALL; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
llama : add support for Orion-14B (#5118) * add support for Orion-14B(https://huggingface.co/OrionStarAI/Orion-14B-Chat) * flake8 support * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * Update llama.cpp * Update llama.cpp --------- Co-authored-by: lixiaopu <lixiaopu@cmcm.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-01-28 08:00:30 +00:00
case LLM_ARCH_ORION:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
llama : add support for Orion-14B (#5118) * add support for Orion-14B(https://huggingface.co/OrionStarAI/Orion-14B-Chat) * flake8 support * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * Update llama.cpp * Update llama.cpp --------- Co-authored-by: lixiaopu <lixiaopu@cmcm.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-01-28 08:00:30 +00:00
switch (hparams.n_layer) {
case 40: model.type = e_model::MODEL_14B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
case LLM_ARCH_INTERNLM2:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
switch (hparams.n_layer) {
case 32: model.type = e_model::MODEL_7B; break;
case 48: model.type = e_model::MODEL_20B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
case LLM_ARCH_GEMMA:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
switch (hparams.n_layer) {
case 18: model.type = e_model::MODEL_2B; break;
case 28: model.type = e_model::MODEL_7B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
llama : add StarCoder2 support (#5795) * Add support for starcoder2 * handle rope type * skip rope freq and rotary embeddings from being serialized * resolve comments * Update llama.cpp * remove redundant changes * handle `rope-theta` * llama : change starcoder2 rope type * address comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-01 19:30:46 +00:00
case LLM_ARCH_STARCODER2:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
switch (hparams.n_layer) {
case 30: model.type = e_model::MODEL_3B; break;
case 32: model.type = e_model::MODEL_7B; break;
case 40: model.type = e_model::MODEL_15B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
case LLM_ARCH_MAMBA:
{
ml.get_key(LLM_KV_SSM_CONV_KERNEL, hparams.ssm_d_conv);
ml.get_key(LLM_KV_SSM_INNER_SIZE, hparams.ssm_d_inner);
ml.get_key(LLM_KV_SSM_STATE_SIZE, hparams.ssm_d_state);
ml.get_key(LLM_KV_SSM_TIME_STEP_RANK, hparams.ssm_dt_rank);
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
switch (hparams.n_layer) {
case 24:
switch (hparams.n_embd) {
case 768: model.type = e_model::MODEL_SMALL; break;
default: model.type = e_model::MODEL_UNKNOWN;
} break;
case 48:
switch (hparams.n_embd) {
case 1024: model.type = e_model::MODEL_MEDIUM; break;
case 1536: model.type = e_model::MODEL_LARGE; break;
case 2048: model.type = e_model::MODEL_XL; break;
default: model.type = e_model::MODEL_UNKNOWN;
} break;
case 64:
switch (hparams.n_embd) {
case 2560: model.type = e_model::MODEL_3B; break;
default: model.type = e_model::MODEL_UNKNOWN;
} break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
[Model] Add support for xverse (#6301) * Support xverse model convert to gguf format. * 1. Convert xverse models to gguf; 2. Add LLM_ARCH_XVERSE inference in llama.cpp; 3. Add xverse item in Supported models in README.md; * * gguf-py: remove redundant logs * llama: remove the init_mapping_prefetch custom parameter * llama.cpp: Include the changes from #6122 to exclude the unused outputs of the last layers. * - Fix format issues - Remove duplicate set kqv_out to llm_build_kv * Update llama.cpp --------- Co-authored-by: willhe <willhe@xverse.cn> Co-authored-by: willhe <hexin@xverse.cn>
2024-03-29 13:37:03 +00:00
case LLM_ARCH_XVERSE:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
switch (hparams.n_layer) {
case 32: model.type = e_model::MODEL_7B; break;
case 40: model.type = e_model::MODEL_13B; break;
case 80: model.type = e_model::MODEL_65B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
case LLM_ARCH_COMMAND_R:
{
ml.get_key(LLM_KV_LOGIT_SCALE, hparams.f_logit_scale);
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
switch (hparams.n_layer) {
case 40: model.type = e_model::MODEL_35B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
case LLM_ARCH_DBRX:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
ml.get_key(LLM_KV_ATTENTION_CLAMP_KQV, hparams.f_clamp_kqv);
switch (hparams.n_layer) {
case 40: model.type = e_model::MODEL_16x12B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
Implement the OLMo architecture (#6741) * implement olmo architecture * remove unused variable * remove unused moe branch * remove check for weight * remove superfluous moe, bias and rope tensors * clarified comment * fix clamp_kqv setting * remove obsolete parameter name filter
2024-04-19 09:35:54 +00:00
case LLM_ARCH_OLMO:
{
ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps);
ml.get_key(LLM_KV_ATTENTION_CLAMP_KQV, hparams.f_clamp_kqv, false);
switch (hparams.n_layer) {
case 22: model.type = e_model::MODEL_1B; break;
case 32: model.type = e_model::MODEL_7B; break;
case 80: model.type = e_model::MODEL_70B; break;
default: model.type = e_model::MODEL_UNKNOWN;
}
} break;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
default: (void)0;
build : enable more non-default compiler warnings (#3200)
2023-09-28 21:41:44 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
model.ftype = ml.ftype;
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
if (hparams.f_max_alibi_bias > 0.0f) {
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
hparams.use_alibi = true;
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
hparams.rope_type = llama_rope_type(&model);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
CUDA full GPU acceleration, KV cache in VRAM (#1827) * Fixed CUDA RoPE * ggml_cuda_mul_mat_vec_p021 * ggml_cuda_scale * ggml_cuda_diag_mask_inf * ggml_is_permuted * ggml_cuda_cpy * flatten rows for ggml_cuda_op * Added a --low-vram option * Fixed Windows performance * Fixed LLAMA_CUDA_DMMV_Y > 1 for WizardLM
2023-06-14 17:47:19 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// TODO: This should probably be in llama.h
BERT tokenizer fixes (#6498) Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
2024-04-09 17:44:08 +00:00
static std::vector<llama_vocab::id> llama_tokenize_internal(
const llama_vocab & vocab, std::string raw_text, bool add_special, bool parse_special = false
);
llama : more tokenizer fixes (#2810) * tests : write a Python tokenizer test (wip) * llama : prefix input text for tokenization with whitespace * llama : distinguish pieces from decoded text + fix detokenization * common : add comments * examples : no longer manually add leading space when tokenizing * tests : use Python to generate tokenizer tests for C++ * tests : add option to tokenize text files ggml-ci * tests : add test-tokenizer-1.py * llama.cpp : fix LF token * hellaswag : move the concat space for clarity * tests : add falcon tests (py + cpp, currently do not pass Unicode) ggml-ci * common : temporary separate llama_detokenize calls for SPM and BPE --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com>
2023-08-27 11:19:19 +00:00
static llama_token llama_byte_to_token(const llama_vocab & vocab, uint8_t ch);
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
static void llm_load_vocab(
llama_model_loader & ml,
llama_model & model) {
auto & vocab = model.vocab;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
struct gguf_context * ctx = ml.meta;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
const auto kv = LLM_KV(model.arch);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// determine vocab type
Print model version. Also improve model type printing, and fix indentation of an unrelated switch statement.
2023-04-08 20:08:21 +00:00
{
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
std::string tokenizer_model;
std::string tokenizer_pre;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
ml.get_key(LLM_KV_TOKENIZER_MODEL, tokenizer_model);
ml.get_key(LLM_KV_TOKENIZER_PRE, tokenizer_pre, false);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
if (tokenizer_model == "no_vocab") {
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
vocab.type = LLAMA_VOCAB_TYPE_NONE;
// default special tokens
BERT tokenizer fixes (#6498) Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
2024-04-09 17:44:08 +00:00
vocab.special_bos_id = -1;
vocab.special_eos_id = -1;
vocab.special_unk_id = -1;
vocab.special_sep_id = -1;
vocab.special_pad_id = -1;
vocab.special_cls_id = -1;
vocab.special_mask_id = -1;
vocab.linefeed_id = -1;
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
return;
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
} else if (tokenizer_model == "llama") {
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
vocab.type = LLAMA_VOCAB_TYPE_SPM;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// default special tokens
BERT tokenizer fixes (#6498) Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
2024-04-09 17:44:08 +00:00
vocab.special_bos_id = 1;
vocab.special_eos_id = 2;
vocab.special_unk_id = 0;
vocab.special_sep_id = -1;
vocab.special_pad_id = -1;
vocab.special_cls_id = -1;
vocab.special_mask_id = -1;
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
gguf : add special tokens metadata for FIM/Infill (#6689) This commit adds special token metadata for Fill-In-the-Middle (FIM)/Infill to the GGUF model. The motivation for this is that currently there is support for CodeLlama but other models exist now like CodeGemma, but the different models use different token ids for the special tokens and this commit allows for supporting multiple models. Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
2024-04-16 06:13:13 +00:00
// For Fill-In-the-Middle (FIM)/infill models which where converted
// prior to support of FIM special tokens in GGUF, the following
// will allow those models to continue to work. The general names
// of the known models are currently CodeLlama (LLM_ARCH_LLAMA) and
// CodeGemma (LLM_ARCH_GEMMA). This can potentially be removed once
// new versions of these models have been published.
std::string gen_name;
llama : make general.name optional (#6709)
2024-04-16 20:50:38 +00:00
ml.get_key(LLM_KV_GENERAL_NAME, gen_name, false);
gguf : add special tokens metadata for FIM/Infill (#6689) This commit adds special token metadata for Fill-In-the-Middle (FIM)/Infill to the GGUF model. The motivation for this is that currently there is support for CodeLlama but other models exist now like CodeGemma, but the different models use different token ids for the special tokens and this commit allows for supporting multiple models. Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
2024-04-16 06:13:13 +00:00
std::transform(gen_name.begin(), gen_name.end(), gen_name.begin(),
[](unsigned char c){ return std::tolower(c); });
llama : make general.name optional (#6709)
2024-04-16 20:50:38 +00:00
gguf : add special tokens metadata for FIM/Infill (#6689) This commit adds special token metadata for Fill-In-the-Middle (FIM)/Infill to the GGUF model. The motivation for this is that currently there is support for CodeLlama but other models exist now like CodeGemma, but the different models use different token ids for the special tokens and this commit allows for supporting multiple models. Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
2024-04-16 06:13:13 +00:00
if (gen_name.find("code") != std::string::npos) {
if (model.arch == LLM_ARCH_LLAMA) {
vocab.special_prefix_id = 32007;
vocab.special_suffix_id = 32008;
vocab.special_middle_id = 32009;
vocab.special_eot_id = 32010;
} else if (model.arch == LLM_ARCH_GEMMA) {
vocab.special_prefix_id = 67;
vocab.special_suffix_id = 69;
vocab.special_middle_id = 68;
llama : support Llama 3 HF conversion (#6745) * Support Llama 3 conversion The tokenizer is BPE. * style * Accept suggestion Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> * llama : add llama_token_is_eog() ggml-ci * llama : auto-detect more EOT tokens when missing in KV data * convert : replacing EOS token is a hack * llama : fix codegemma EOT token + add TODOs * llama : fix model type string for 8B model --------- Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-21 11:50:41 +00:00
// TODO: this is not EOT, it is "file separator" token, needs fix
// https://huggingface.co/google/codegemma-7b-it/blob/9b1d9231388358c04d90bd003458f5070d97db44/tokenizer_config.json#L565-L572
//vocab.special_eot_id = 70;
vocab.special_eot_id = 107;
gguf : add special tokens metadata for FIM/Infill (#6689) This commit adds special token metadata for Fill-In-the-Middle (FIM)/Infill to the GGUF model. The motivation for this is that currently there is support for CodeLlama but other models exist now like CodeGemma, but the different models use different token ids for the special tokens and this commit allows for supporting multiple models. Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
2024-04-16 06:13:13 +00:00
}
}
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
const int add_space_prefix_keyidx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_ADD_PREFIX).c_str());
if (add_space_prefix_keyidx != -1) {
vocab.add_space_prefix = gguf_get_val_bool(ctx, add_space_prefix_keyidx);
} // The default value of add_space_prefix is true.
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
} else if (tokenizer_model == "bert") {
vocab.type = LLAMA_VOCAB_TYPE_WPM;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
// default special tokens
vocab.special_bos_id = -1;
vocab.special_eos_id = -1;
vocab.special_unk_id = 100;
vocab.special_sep_id = 102;
vocab.special_pad_id = 0;
vocab.special_cls_id = 101;
vocab.special_mask_id = 103;
vocab.add_space_prefix = false;
} else {
if (tokenizer_model == "gpt2") {
vocab.type = LLAMA_VOCAB_TYPE_BPE;
} else {
LLAMA_LOG_WARN("%s: unknown tokenizer: '%s'", __func__, tokenizer_model.c_str());
LLAMA_LOG_WARN("%s: using default tokenizer: 'llama'", __func__);
vocab.type = LLAMA_VOCAB_TYPE_SPM;
return;
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// read bpe merges and populate bpe ranks
const int merges_keyidx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_MERGES).c_str());
if (merges_keyidx == -1) {
throw std::runtime_error("cannot find tokenizer merges in model file\n");
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
const int n_merges = gguf_get_arr_n(ctx, merges_keyidx);
ggml : use F16 instead of F32 in Q4_0, Q4_1, Q8_0 (#1508) * ggml : use F16 instead of F32 in Q4_0, Q4_1 and Q8_0 * llama : bump LLAMA_FILE_VERSION to 3 * cuda : update Q4 and Q8 dequantize kernels * ggml : fix AVX dot products * readme : update performance table + hot topics
2023-05-19 19:17:18 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
for (int i = 0; i < n_merges; i++) {
const std::string word = gguf_get_arr_str(ctx, merges_keyidx, i);
llama : refactor unicode stuff (#5992) * llama : refactor unicode stuff ggml-ci * unicode : names * make : fix c++ compiler * unicode : names * unicode : straighten tables * zig : fix build * unicode : put nfd normalization behind API ggml-ci * swift : fix build * unicode : add BOM * unicode : add <cstdint> ggml-ci * unicode : pass as cpts as const ref
2024-03-11 15:47:47 +00:00
GGML_ASSERT(unicode_cpts_from_utf8(word).size() > 0);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
std::string first;
std::string second;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
const size_t pos = word.find(' ', 1);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
if (pos != std::string::npos) {
first = word.substr(0, pos);
second = word.substr(pos + 1);
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
vocab.bpe_ranks.emplace(std::make_pair(first, second), i);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// default special tokens
BERT tokenizer fixes (#6498) Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
2024-04-09 17:44:08 +00:00
vocab.special_bos_id = 11;
vocab.special_eos_id = 11;
vocab.special_unk_id = -1;
vocab.special_sep_id = -1;
vocab.special_pad_id = -1;
vocab.special_cls_id = -1;
vocab.special_mask_id = -1;
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
}
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
// for now, only BPE models have pre-tokenizers
if (vocab.type == LLAMA_VOCAB_TYPE_BPE) {
if (tokenizer_pre.empty()) {
LLAMA_LOG_WARN("%s: missing pre-tokenizer type, using: 'default'\n", __func__);
LLAMA_LOG_WARN("%s: \n", __func__);
LLAMA_LOG_WARN("%s: ************************************ \n", __func__);
LLAMA_LOG_WARN("%s: GENERATION QUALITY WILL BE DEGRADED! \n", __func__);
LLAMA_LOG_WARN("%s: CONSIDER REGENERATING THE MODEL \n", __func__);
LLAMA_LOG_WARN("%s: ************************************ \n", __func__);
LLAMA_LOG_WARN("%s: \n", __func__);
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
} else if (
tokenizer_pre == "default") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
} else if (
tokenizer_pre == "llama3" ||
tokenizer_pre == "llama-v3" ||
tokenizer_pre == "llama-bpe") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_LLAMA3;
} else if (
tokenizer_pre == "deepseek-llm") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEEPSEEK_LLM;
} else if (
tokenizer_pre == "deepseek-coder") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEEPSEEK_CODER;
} else if (
tokenizer_pre == "falcon") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_FALCON;
} else if (
tokenizer_pre == "mpt") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_MPT;
} else if (
tokenizer_pre == "starcoder") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_STARCODER;
} else if (
tokenizer_pre == "gpt-2") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_GPT2;
tests : add test-tokenizer-0.sh + fix some tokenizers (#7036) * tests : add test-tokenizer-0.sh * unicode : add all unicode number ranges * starcoder : fix pre-tokenizer * tests : add test that fails with DeepSeek tokenizers * falcon : fix regex * unicode : regenerate unicode tables * refact : add tokenizer model * lint : fix * tests : disable failing tests ggml-ci * refact : add tests files ggml-ci * convert : print -> logging ggml-ci * lint : fix * unicode : digit -> number * phi-3 : update
2024-05-04 05:32:32 +00:00
} else if (
tokenizer_pre == "refact") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_REFACT;
command-r : add BPE pre-tokenization (#7063) * Add BPE pre-tokenization for Command-R/R+. * Bump transformers convert requirement. * command-r : add individual digits regex --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-05-05 05:19:30 +00:00
} else if (
tokenizer_pre == "command-r") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_COMMAND_R;
Fix OLMo HF to GGUF conversion (#6910)
2024-05-07 19:39:43 +00:00
} else if (
tokenizer_pre == "olmo") {
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_OLMO;
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
} else {
throw std::runtime_error(format("unknown pre-tokenizer type: '%s'", tokenizer_pre.c_str()));
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
} else {
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT;
ggml : remove bit shuffling (#1405) * ggml : remove Q4_0 bit shufling (ARM NEON) * ggml : remove Q4_1 bit shuffling (ARM NEON + reference) * ggml : nibbles_from_floats() + bytes_from_nibbles() (ARM NEON) * ggml : remove Q4_2 bit shuffling (WIP, BROKEN) * ggml : remove Q5_0 bit shuffling (ARM NEON) * ggml : 2x faster scalar implementations * ggml : remove Q5_1 bit shuffling (ARM NEON + scalar) * ggml : simplify scalar dot * ggml : remove WASM SIMD bit shuffling + remove vzip for ARM 32-bit * ggml : fix Q4_1 quantization * ggml : update cuBLAS + normalize variable names * ggml : remove Q4_2 mode * ggml : minor formatting * ggml : fix Q5_0 quantization * scripts : add script for measuring the time per token * AVX implementations (#1370) * ggml : uniform 5th bit extraction * llama : produce error upon loading old model files * llama : fix model magic/version write * ggml : speed-up Q5_0 + Q5_1 at 4 threads * ggml : preserve old Q4 and Q5 formats * ggml : simplify Q8_1 - no need for low / high sums anymore * ggml : fix Q8_0 and Q8_1 rounding * Revert "AVX implementations (#1370)" This reverts commit 948d124837f9d287d8490f41338e0e4cceb0814f. * ggml : fix AVX2 implementation * sha : update hashes for 7B and 13B * readme : update timings + remove warning banner * llama : update v2 PR number to 1405 * ggml : fix WASM comments * ggml : back to original bit order * readme : add note that Q4 and Q5 have been changed * llama : fix return for unknown version --------- Co-authored-by: Stephan Walter <stephan@walter.name>
2023-05-11 21:23:08 +00:00
}
}
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
const int token_idx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_LIST).c_str());
if (token_idx == -1) {
throw std::runtime_error("cannot find tokenizer vocab in model file\n");
}
const float * scores = nullptr;
const int score_idx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_SCORES).c_str());
if (score_idx != -1) {
scores = (const float * ) gguf_get_arr_data(ctx, score_idx);
}
const int * toktypes = nullptr;
const int toktype_idx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_TOKEN_TYPE).c_str());
if (toktype_idx != -1) {
toktypes = (const int * ) gguf_get_arr_data(ctx, toktype_idx);
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
const uint32_t n_vocab = gguf_get_arr_n(ctx, token_idx);
vocab.id_to_token.resize(n_vocab);
for (uint32_t i = 0; i < n_vocab; i++) {
std::string word = gguf_get_arr_str(ctx, token_idx, i);
llama : refactor unicode stuff (#5992) * llama : refactor unicode stuff ggml-ci * unicode : names * make : fix c++ compiler * unicode : names * unicode : straighten tables * zig : fix build * unicode : put nfd normalization behind API ggml-ci * swift : fix build * unicode : add BOM * unicode : add <cstdint> ggml-ci * unicode : pass as cpts as const ref
2024-03-11 15:47:47 +00:00
GGML_ASSERT(unicode_cpts_from_utf8(word).size() > 0);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
vocab.token_to_id[word] = i;
auto & token_data = vocab.id_to_token[i];
token_data.text = std::move(word);
gguf : make token scores and types optional (#3347)
2023-09-28 18:30:15 +00:00
token_data.score = scores ? scores[i] : 0.0f;
token_data.type = toktypes ? (llama_token_type) toktypes[i] : LLAMA_TOKEN_TYPE_NORMAL;
Add mmap support for model files
2023-03-29 00:03:43 +00:00
}
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
GGML_ASSERT(vocab.id_to_token.size() == vocab.token_to_id.size());
Add mmap support for model files
2023-03-29 00:03:43 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// determine the newline token: LLaMA "<0x0A>" == 10 == '\n', Falcon 193 == '\n'
llama : more tokenizer fixes (#2810) * tests : write a Python tokenizer test (wip) * llama : prefix input text for tokenization with whitespace * llama : distinguish pieces from decoded text + fix detokenization * common : add comments * examples : no longer manually add leading space when tokenizing * tests : use Python to generate tokenizer tests for C++ * tests : add option to tokenize text files ggml-ci * tests : add test-tokenizer-1.py * llama.cpp : fix LF token * hellaswag : move the concat space for clarity * tests : add falcon tests (py + cpp, currently do not pass Unicode) ggml-ci * common : temporary separate llama_detokenize calls for SPM and BPE --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com>
2023-08-27 11:19:19 +00:00
if (vocab.type == LLAMA_VOCAB_TYPE_SPM) {
llama : allow raw byte in SPM vocabs; don't crash on nl 404 (#5478) * common : don't crash if newline token is not found * common : llama_byte_to_token: allow falling back to finding just the token byte in SPM vocabs
2024-02-13 16:18:16 +00:00
try {
vocab.linefeed_id = llama_byte_to_token(vocab, '\n');
} catch (const std::exception & e) {
LLAMA_LOG_WARN("%s: SPM vocabulary, but newline token not found: %s! Using special_pad_id instead.", __func__, e.what());
vocab.linefeed_id = vocab.special_pad_id;
}
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
} else if (vocab.type == LLAMA_VOCAB_TYPE_WPM) {
vocab.linefeed_id = vocab.special_pad_id;
llama : more tokenizer fixes (#2810) * tests : write a Python tokenizer test (wip) * llama : prefix input text for tokenization with whitespace * llama : distinguish pieces from decoded text + fix detokenization * common : add comments * examples : no longer manually add leading space when tokenizing * tests : use Python to generate tokenizer tests for C++ * tests : add option to tokenize text files ggml-ci * tests : add test-tokenizer-1.py * llama.cpp : fix LF token * hellaswag : move the concat space for clarity * tests : add falcon tests (py + cpp, currently do not pass Unicode) ggml-ci * common : temporary separate llama_detokenize calls for SPM and BPE --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com>
2023-08-27 11:19:19 +00:00
} else {
use _wfopen instead of fopen on Windows (#6248) also fix missing #defines before windows.h, and BPE LF token on MSVC
2024-03-23 22:48:02 +00:00
const std::vector<int> ids = llama_tokenize_internal(vocab, "\xC4\x8A", false); // U+010A
llama : validate special token ids are in range when loading GGUF model (#3635) * Add validation for special token ids to llama.cpp Small optimization for llama_byte_to_token SPM mode * Fix BPE newline check, only I could break something so simple * Killll meeeeee * Account for GGUF_KEY_KEY only setting when the key exists * Minor code cleanups. * Fix convert.py error msg when added tokens are out of range * Make gguf SpecialVocab vocab size-aware Update conversion scripts accordingly * Avoid a string copy Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-22 18:14:56 +00:00
GGML_ASSERT(!ids.empty() && "model vocab missing newline token");
vocab.linefeed_id = ids[0];
llama : more tokenizer fixes (#2810) * tests : write a Python tokenizer test (wip) * llama : prefix input text for tokenization with whitespace * llama : distinguish pieces from decoded text + fix detokenization * common : add comments * examples : no longer manually add leading space when tokenizing * tests : use Python to generate tokenizer tests for C++ * tests : add option to tokenize text files ggml-ci * tests : add test-tokenizer-1.py * llama.cpp : fix LF token * hellaswag : move the concat space for clarity * tests : add falcon tests (py + cpp, currently do not pass Unicode) ggml-ci * common : temporary separate llama_detokenize calls for SPM and BPE --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com>
2023-08-27 11:19:19 +00:00
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// special tokens
llama : validate special token ids are in range when loading GGUF model (#3635) * Add validation for special token ids to llama.cpp Small optimization for llama_byte_to_token SPM mode * Fix BPE newline check, only I could break something so simple * Killll meeeeee * Account for GGUF_KEY_KEY only setting when the key exists * Minor code cleanups. * Fix convert.py error msg when added tokens are out of range * Make gguf SpecialVocab vocab size-aware Update conversion scripts accordingly * Avoid a string copy Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-22 18:14:56 +00:00
{
const std::vector<std::pair<enum llm_kv, int32_t &>> special_token_types = {
gguf : add special tokens metadata for FIM/Infill (#6689) This commit adds special token metadata for Fill-In-the-Middle (FIM)/Infill to the GGUF model. The motivation for this is that currently there is support for CodeLlama but other models exist now like CodeGemma, but the different models use different token ids for the special tokens and this commit allows for supporting multiple models. Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
2024-04-16 06:13:13 +00:00
{ LLM_KV_TOKENIZER_BOS_ID, vocab.special_bos_id },
{ LLM_KV_TOKENIZER_EOS_ID, vocab.special_eos_id },
{ LLM_KV_TOKENIZER_UNK_ID, vocab.special_unk_id },
{ LLM_KV_TOKENIZER_SEP_ID, vocab.special_sep_id },
{ LLM_KV_TOKENIZER_PAD_ID, vocab.special_pad_id },
{ LLM_KV_TOKENIZER_CLS_ID, vocab.special_cls_id },
{ LLM_KV_TOKENIZER_MASK_ID, vocab.special_mask_id },
{ LLM_KV_TOKENIZER_PREFIX_ID, vocab.special_prefix_id },
{ LLM_KV_TOKENIZER_SUFFIX_ID, vocab.special_suffix_id },
{ LLM_KV_TOKENIZER_MIDDLE_ID, vocab.special_middle_id },
{ LLM_KV_TOKENIZER_EOT_ID, vocab.special_eot_id },
llama : validate special token ids are in range when loading GGUF model (#3635) * Add validation for special token ids to llama.cpp Small optimization for llama_byte_to_token SPM mode * Fix BPE newline check, only I could break something so simple * Killll meeeeee * Account for GGUF_KEY_KEY only setting when the key exists * Minor code cleanups. * Fix convert.py error msg when added tokens are out of range * Make gguf SpecialVocab vocab size-aware Update conversion scripts accordingly * Avoid a string copy Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-22 18:14:56 +00:00
};
llama : support Llama 3 HF conversion (#6745) * Support Llama 3 conversion The tokenizer is BPE. * style * Accept suggestion Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> * llama : add llama_token_is_eog() ggml-ci * llama : auto-detect more EOT tokens when missing in KV data * convert : replacing EOS token is a hack * llama : fix codegemma EOT token + add TODOs * llama : fix model type string for 8B model --------- Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-21 11:50:41 +00:00
llama : validate special token ids are in range when loading GGUF model (#3635) * Add validation for special token ids to llama.cpp Small optimization for llama_byte_to_token SPM mode * Fix BPE newline check, only I could break something so simple * Killll meeeeee * Account for GGUF_KEY_KEY only setting when the key exists * Minor code cleanups. * Fix convert.py error msg when added tokens are out of range * Make gguf SpecialVocab vocab size-aware Update conversion scripts accordingly * Avoid a string copy Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-22 18:14:56 +00:00
for (const auto & it : special_token_types) {
const std::string & key = kv(std::get<0>(it));
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
int32_t & id = std::get<1>(it);
llama : validate special token ids are in range when loading GGUF model (#3635) * Add validation for special token ids to llama.cpp Small optimization for llama_byte_to_token SPM mode * Fix BPE newline check, only I could break something so simple * Killll meeeeee * Account for GGUF_KEY_KEY only setting when the key exists * Minor code cleanups. * Fix convert.py error msg when added tokens are out of range * Make gguf SpecialVocab vocab size-aware Update conversion scripts accordingly * Avoid a string copy Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-22 18:14:56 +00:00
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
uint32_t new_id;
if (!ml.get_key(std::get<0>(it), new_id, false)) {
continue;
}
if (new_id >= vocab.id_to_token.size()) {
LLAMA_LOG_WARN("%s: bad special token: '%s' = %ud, using default id %d\n",
__func__, key.c_str(), new_id, id);
} else {
id = new_id;
llama : validate special token ids are in range when loading GGUF model (#3635) * Add validation for special token ids to llama.cpp Small optimization for llama_byte_to_token SPM mode * Fix BPE newline check, only I could break something so simple * Killll meeeeee * Account for GGUF_KEY_KEY only setting when the key exists * Minor code cleanups. * Fix convert.py error msg when added tokens are out of range * Make gguf SpecialVocab vocab size-aware Update conversion scripts accordingly * Avoid a string copy Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-22 18:14:56 +00:00
}
Respect tokenizer.ggml.add_bos_token value when tokenizing (#4040) * gguf-py: gguf-dump: Respect --no-tensor flag in JSON mode. * Respect add_bos_token GGUF metadata value * gguf-py: Try to fix SpecialVocab giving up too easily for the Nth time
2023-11-17 02:14:37 +00:00
}
// Handle add_bos_token and add_eos_token
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
{
bool temp = true;
if (ml.get_key(LLM_KV_TOKENIZER_ADD_BOS, temp, false)) {
vocab.special_add_bos = int(temp);
}
if (ml.get_key(LLM_KV_TOKENIZER_ADD_EOS, temp, false)) {
vocab.special_add_eos = int(temp);
}
llama : validate special token ids are in range when loading GGUF model (#3635) * Add validation for special token ids to llama.cpp Small optimization for llama_byte_to_token SPM mode * Fix BPE newline check, only I could break something so simple * Killll meeeeee * Account for GGUF_KEY_KEY only setting when the key exists * Minor code cleanups. * Fix convert.py error msg when added tokens are out of range * Make gguf SpecialVocab vocab size-aware Update conversion scripts accordingly * Avoid a string copy Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-22 18:14:56 +00:00
}
llama : support Llama 3 HF conversion (#6745) * Support Llama 3 conversion The tokenizer is BPE. * style * Accept suggestion Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> * llama : add llama_token_is_eog() ggml-ci * llama : auto-detect more EOT tokens when missing in KV data * convert : replacing EOS token is a hack * llama : fix codegemma EOT token + add TODOs * llama : fix model type string for 8B model --------- Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-21 11:50:41 +00:00
llama : fix typo in <|im_end|> token text (#6745)
2024-04-22 12:41:11 +00:00
// find EOT token: "<|eot_id|>", "<|im_end|>", "<end_of_turn>", etc.
llama : support Llama 3 HF conversion (#6745) * Support Llama 3 conversion The tokenizer is BPE. * style * Accept suggestion Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> * llama : add llama_token_is_eog() ggml-ci * llama : auto-detect more EOT tokens when missing in KV data * convert : replacing EOS token is a hack * llama : fix codegemma EOT token + add TODOs * llama : fix model type string for 8B model --------- Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-21 11:50:41 +00:00
//
// TODO: convert scripts should provide this token through the KV metadata LLAMA_KV_TOKENIZER_EOT_ID
// for now, we apply this workaround to find the EOT token based on its text
if (vocab.special_eot_id == -1) {
for (const auto & t : vocab.token_to_id) {
if (
// TODO: gemma "<end_of_turn>" is exported as a normal token, so the following check does not work
// need to fix convert script
//vocab.id_to_token[t.second].type == LLAMA_TOKEN_TYPE_CONTROL &&
(t.first == "<|eot_id|>" ||
llama : fix typo in <|im_end|> token text (#6745)
2024-04-22 12:41:11 +00:00
t.first == "<|im_end|>" ||
llama : add phi3 support (#6852) * add explicit phi3 support * add explicit phi3 support * remove unused code * convert : add BOS token * llama : match EOT token <|end|> * llama : minor / style * llama : tabs -> spaces * convert : fix lint checks --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-24 07:00:37 +00:00
t.first == "<|end|>" ||
llama : support Llama 3 HF conversion (#6745) * Support Llama 3 conversion The tokenizer is BPE. * style * Accept suggestion Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> * llama : add llama_token_is_eog() ggml-ci * llama : auto-detect more EOT tokens when missing in KV data * convert : replacing EOS token is a hack * llama : fix codegemma EOT token + add TODOs * llama : fix model type string for 8B model --------- Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-21 11:50:41 +00:00
t.first == "<end_of_turn>"
)
) {
vocab.special_eot_id = t.second;
break;
}
}
}
llama : validate special token ids are in range when loading GGUF model (#3635) * Add validation for special token ids to llama.cpp Small optimization for llama_byte_to_token SPM mode * Fix BPE newline check, only I could break something so simple * Killll meeeeee * Account for GGUF_KEY_KEY only setting when the key exists * Minor code cleanups. * Fix convert.py error msg when added tokens are out of range * Make gguf SpecialVocab vocab size-aware Update conversion scripts accordingly * Avoid a string copy Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-22 18:14:56 +00:00
}
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
// build special tokens cache
{
// TODO: It is unclear (to me) at this point, whether special tokes are guaranteed to be of a deterministic type,
// and will always be correctly labeled in 'added_tokens.json' etc.
// The assumption is, since special tokens aren't meant to be exposed to end user, they are designed
// to be unmatchable by the tokenizer, therefore tokens from the vocab, which are unmatchable by the tokenizer
// are special tokens.
english : use `typos` to fix comments and logs (#4354)
2023-12-12 09:53:36 +00:00
// From testing, this appears to correlate 1:1 with special tokens.
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
//
// Counting special tokens and verifying in only one direction
// is sufficient to detect difference in those two sets.
//
uint32_t special_tokens_count_by_type = 0;
uint32_t special_tokens_count_from_verification = 0;
bool special_tokens_definition_mismatch = false;
for (const auto & t : vocab.token_to_id) {
const auto & token = t.first;
const auto & id = t.second;
// Count all non-normal tokens in the vocab while iterating
if (vocab.id_to_token[id].type != LLAMA_TOKEN_TYPE_NORMAL) {
special_tokens_count_by_type++;
}
// Skip single character tokens
if (token.length() > 1) {
bool is_tokenizable = false;
// Split token string representation in two, in all possible ways
// and check if both halves can be matched to a valid token
for (unsigned i = 1; i < token.length();) {
const auto left = token.substr(0, i);
const auto right = token.substr(i);
// check if we didnt partition in the middle of a utf sequence
auto utf = utf8_len(left.at(left.length() - 1));
if (utf == 1) {
if (vocab.token_to_id.find(left) != vocab.token_to_id.end() &&
vocab.token_to_id.find(right) != vocab.token_to_id.end() ) {
is_tokenizable = true;
break;
}
i++;
} else {
// skip over the rest of multibyte utf sequence
i += utf - 1;
}
}
if (!is_tokenizable) {
// Some tokens are multibyte, but they are utf sequences with equivalent text length of 1
// it's faster to re-filter them here, since there are way less candidates now
// Calculate a total "utf" length of a token string representation
size_t utf8_str_len = 0;
for (unsigned i = 0; i < token.length();) {
utf8_str_len++;
i += utf8_len(token.at(i));
}
// And skip the ones which are one character
if (utf8_str_len > 1) {
// At this point what we have left are special tokens only
vocab.special_tokens_cache[token] = id;
// Count manually found special tokens
special_tokens_count_from_verification++;
// If this manually found special token is not marked as such, flag a mismatch
if (vocab.id_to_token[id].type == LLAMA_TOKEN_TYPE_NORMAL) {
special_tokens_definition_mismatch = true;
}
}
}
}
}
if (special_tokens_definition_mismatch || special_tokens_count_from_verification != special_tokens_count_by_type) {
llama : avoid fprintf in favor of LLAMA_LOG (#3538)
2023-10-17 19:34:26 +00:00
LLAMA_LOG_WARN("%s: mismatch in special tokens definition ( %u/%zu vs %u/%zu ).\n",
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
__func__,
special_tokens_count_from_verification, vocab.id_to_token.size(),
special_tokens_count_by_type, vocab.id_to_token.size()
);
} else {
llama : avoid fprintf in favor of LLAMA_LOG (#3538)
2023-10-17 19:34:26 +00:00
LLAMA_LOG_INFO("%s: special tokens definition check successful ( %u/%zu ).\n",
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
__func__,
special_tokens_count_from_verification, vocab.id_to_token.size()
);
}
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
static void llm_load_print_meta(llama_model_loader & ml, llama_model & model) {
const auto & hparams = model.hparams;
const auto & vocab = model.vocab;
YaRN : store rope scaling type as int32_t in memory (#5285) * YaRN : store rope scaling type as int32_t in memory * llama : store mapped names as const char *
2024-02-03 11:22:06 +00:00
const char * rope_scaling_type = LLAMA_ROPE_SCALING_TYPES.at(hparams.rope_scaling_type_train);
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// hparams
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
LLAMA_LOG_INFO("%s: format = %s\n", __func__, llama_file_version_name(ml.fver));
YaRN : store rope scaling type as int32_t in memory (#5285) * YaRN : store rope scaling type as int32_t in memory * llama : store mapped names as const char *
2024-02-03 11:22:06 +00:00
LLAMA_LOG_INFO("%s: arch = %s\n", __func__, LLM_ARCH_NAMES.at(model.arch));
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
LLAMA_LOG_INFO("%s: vocab type = %s\n", __func__, llama_model_vocab_type_name(vocab.type));
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
LLAMA_LOG_INFO("%s: n_vocab = %u\n", __func__, hparams.n_vocab);
LLAMA_LOG_INFO("%s: n_merges = %u\n", __func__, (int) vocab.bpe_ranks.size());
LLAMA_LOG_INFO("%s: n_ctx_train = %u\n", __func__, hparams.n_ctx_train);
LLAMA_LOG_INFO("%s: n_embd = %u\n", __func__, hparams.n_embd);
LLAMA_LOG_INFO("%s: n_head = %u\n", __func__, hparams.n_head);
LLAMA_LOG_INFO("%s: n_head_kv = %u\n", __func__, hparams.n_head_kv);
LLAMA_LOG_INFO("%s: n_layer = %u\n", __func__, hparams.n_layer);
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
LLAMA_LOG_INFO("%s: n_rot = %u\n", __func__, hparams.n_rot);
LLAMA_LOG_INFO("%s: n_embd_head_k = %u\n", __func__, hparams.n_embd_head_k);
LLAMA_LOG_INFO("%s: n_embd_head_v = %u\n", __func__, hparams.n_embd_head_v);
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
LLAMA_LOG_INFO("%s: n_gqa = %u\n", __func__, hparams.n_gqa());
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
LLAMA_LOG_INFO("%s: n_embd_k_gqa = %u\n", __func__, hparams.n_embd_k_gqa());
LLAMA_LOG_INFO("%s: n_embd_v_gqa = %u\n", __func__, hparams.n_embd_v_gqa());
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
LLAMA_LOG_INFO("%s: f_norm_eps = %.1e\n", __func__, hparams.f_norm_eps);
LLAMA_LOG_INFO("%s: f_norm_rms_eps = %.1e\n", __func__, hparams.f_norm_rms_eps);
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
LLAMA_LOG_INFO("%s: f_clamp_kqv = %.1e\n", __func__, hparams.f_clamp_kqv);
LLAMA_LOG_INFO("%s: f_max_alibi_bias = %.1e\n", __func__, hparams.f_max_alibi_bias);
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
LLAMA_LOG_INFO("%s: f_logit_scale = %.1e\n", __func__, hparams.f_logit_scale);
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
LLAMA_LOG_INFO("%s: n_ff = %u\n", __func__, hparams.n_ff);
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
LLAMA_LOG_INFO("%s: n_expert = %u\n", __func__, hparams.n_expert);
LLAMA_LOG_INFO("%s: n_expert_used = %u\n", __func__, hparams.n_expert_used);
llama : fix typo
2024-03-14 11:13:06 +00:00
LLAMA_LOG_INFO("%s: causal attn = %d\n", __func__, hparams.causal_attn);
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
LLAMA_LOG_INFO("%s: pooling type = %d\n", __func__, hparams.pooling_type);
LLAMA_LOG_INFO("%s: rope type = %d\n", __func__, hparams.rope_type);
YaRN : store rope scaling type as int32_t in memory (#5285) * YaRN : store rope scaling type as int32_t in memory * llama : store mapped names as const char *
2024-02-03 11:22:06 +00:00
LLAMA_LOG_INFO("%s: rope scaling = %s\n", __func__, rope_scaling_type);
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
LLAMA_LOG_INFO("%s: freq_base_train = %.1f\n", __func__, hparams.rope_freq_base_train);
LLAMA_LOG_INFO("%s: freq_scale_train = %g\n", __func__, hparams.rope_freq_scale_train);
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
LLAMA_LOG_INFO("%s: n_yarn_orig_ctx = %u\n", __func__, hparams.n_yarn_orig_ctx);
LLAMA_LOG_INFO("%s: rope_finetuned = %s\n", __func__, hparams.rope_finetuned ? "yes" : "unknown");
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
LLAMA_LOG_INFO("%s: ssm_d_conv = %u\n", __func__, hparams.ssm_d_conv);
LLAMA_LOG_INFO("%s: ssm_d_inner = %u\n", __func__, hparams.ssm_d_inner);
LLAMA_LOG_INFO("%s: ssm_d_state = %u\n", __func__, hparams.ssm_d_state);
LLAMA_LOG_INFO("%s: ssm_dt_rank = %u\n", __func__, hparams.ssm_dt_rank);
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
LLAMA_LOG_INFO("%s: model type = %s\n", __func__, llama_model_type_name(model.type));
LLAMA_LOG_INFO("%s: model ftype = %s\n", __func__, llama_model_ftype_name(model.ftype).c_str());
llama : add additional suffixes for model params (#4834) * llm_load_print_meta: Add additional suffixs for model params * Update llama.cpp model param log remove unneeded comments and convert from > to >=
2024-01-10 14:09:53 +00:00
if (ml.n_elements >= 1e12) {
LLAMA_LOG_INFO("%s: model params = %.2f T\n", __func__, ml.n_elements*1e-12);
} else if (ml.n_elements >= 1e9) {
LLAMA_LOG_INFO("%s: model params = %.2f B\n", __func__, ml.n_elements*1e-9);
} else if (ml.n_elements >= 1e6) {
LLAMA_LOG_INFO("%s: model params = %.2f M\n", __func__, ml.n_elements*1e-6);
} else {
LLAMA_LOG_INFO("%s: model params = %.2f K\n", __func__, ml.n_elements*1e-3);
}
llama : fix data units (#4101) * llama : fix data units ggml-ci * Revert "llama : fix data units" This reverts commit f5feac831fe225ed7f3db938d115732a49dccfc4. * llama : disambiguate data units ggml-ci
2023-11-17 08:00:15 +00:00
if (ml.n_bytes < GiB) {
LLAMA_LOG_INFO("%s: model size = %.2f MiB (%.2f BPW) \n", __func__, ml.n_bytes/1024.0/1024.0, ml.n_bytes*8.0/ml.n_elements);
llama.cpp : show model size and BPW on load (#3223)
2023-09-17 12:33:28 +00:00
} else {
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
LLAMA_LOG_INFO("%s: model size = %.2f GiB (%.2f BPW) \n", __func__, ml.n_bytes/1024.0/1024.0/1024.0, ml.n_bytes*8.0/ml.n_elements);
llama.cpp : show model size and BPW on load (#3223)
2023-09-17 12:33:28 +00:00
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// general kv
llama : fix alignment of general.name in print meta (#4254) * llama: fix alignment of general.name in print meta This commit fixes the alignment of the general.name field in the llm_load_print_meta function. Currently the output looks like this: ```console llm_load_print_meta: model ftype = mostly Q4_0 llm_load_print_meta: model params = 13.02 B llm_load_print_meta: model size = 6.86 GiB (4.53 BPW) llm_load_print_meta: general.name = LLaMA v2 ``` And with this commit it looks like this: ```console llm_load_print_meta: model ftype = mostly Q4_0 llm_load_print_meta: model params = 13.02 B llm_load_print_meta: model size = 6.86 GiB (4.53 BPW) llm_load_print_meta: general.name = LLaMA v2 ``` Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com> * llama: fix alignment of special tokens Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com> --------- Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
2023-11-30 21:43:08 +00:00
LLAMA_LOG_INFO("%s: general.name = %s\n", __func__, model.name.c_str());
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// special tokens
llama : support Llama 3 HF conversion (#6745) * Support Llama 3 conversion The tokenizer is BPE. * style * Accept suggestion Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> * llama : add llama_token_is_eog() ggml-ci * llama : auto-detect more EOT tokens when missing in KV data * convert : replacing EOS token is a hack * llama : fix codegemma EOT token + add TODOs * llama : fix model type string for 8B model --------- Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-21 11:50:41 +00:00
if (vocab.special_bos_id != -1) { LLAMA_LOG_INFO( "%s: BOS token = %d '%s'\n", __func__, vocab.special_bos_id, vocab.id_to_token[vocab.special_bos_id].text.c_str() ); }
if (vocab.special_eos_id != -1) { LLAMA_LOG_INFO( "%s: EOS token = %d '%s'\n", __func__, vocab.special_eos_id, vocab.id_to_token[vocab.special_eos_id].text.c_str() ); }
if (vocab.special_unk_id != -1) { LLAMA_LOG_INFO( "%s: UNK token = %d '%s'\n", __func__, vocab.special_unk_id, vocab.id_to_token[vocab.special_unk_id].text.c_str() ); }
if (vocab.special_sep_id != -1) { LLAMA_LOG_INFO( "%s: SEP token = %d '%s'\n", __func__, vocab.special_sep_id, vocab.id_to_token[vocab.special_sep_id].text.c_str() ); }
if (vocab.special_pad_id != -1) { LLAMA_LOG_INFO( "%s: PAD token = %d '%s'\n", __func__, vocab.special_pad_id, vocab.id_to_token[vocab.special_pad_id].text.c_str() ); }
if (vocab.special_cls_id != -1) { LLAMA_LOG_INFO( "%s: CLS token = %d '%s'\n", __func__, vocab.special_cls_id, vocab.id_to_token[vocab.special_cls_id].text.c_str() ); }
if (vocab.special_mask_id != -1) { LLAMA_LOG_INFO( "%s: MASK token = %d '%s'\n", __func__, vocab.special_mask_id, vocab.id_to_token[vocab.special_mask_id].text.c_str() ); }
if (vocab.linefeed_id != -1) { LLAMA_LOG_INFO( "%s: LF token = %d '%s'\n", __func__, vocab.linefeed_id, vocab.id_to_token[vocab.linefeed_id].text.c_str() ); }
if (vocab.special_prefix_id != -1) { LLAMA_LOG_INFO( "%s: PRE token = %d '%s'\n", __func__, vocab.special_prefix_id, vocab.id_to_token[vocab.special_prefix_id].text.c_str() ); }
if (vocab.special_suffix_id != -1) { LLAMA_LOG_INFO( "%s: SUF token = %d '%s'\n", __func__, vocab.special_suffix_id, vocab.id_to_token[vocab.special_suffix_id].text.c_str() ); }
if (vocab.special_middle_id != -1) { LLAMA_LOG_INFO( "%s: MID token = %d '%s'\n", __func__, vocab.special_middle_id, vocab.id_to_token[vocab.special_middle_id].text.c_str() ); }
if (vocab.special_eot_id != -1) { LLAMA_LOG_INFO( "%s: EOT token = %d '%s'\n", __func__, vocab.special_eot_id, vocab.id_to_token[vocab.special_eot_id].text.c_str() ); }
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
llama : add ability to cancel model loading (#4462) * llama : Add ability to cancel model load Updated llama_progress_callback so that if it returns false, the model loading is aborted. * llama : Add test for model load cancellation * Fix bool return in llama_model_load, remove std::ignore use * Update llama.cpp Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Fail test if model file is missing * Revert "Fail test if model file is missing" This reverts commit 32ebd525bf7e5a87ee8a3dbaab3d92ce79fbf23d. * Add test-model-load-cancel to Makefile * Revert "Revert "Fail test if model file is missing"" This reverts commit 2796953257ee5383fa7c8fe8fa8fc888c048fb0b. * Simplify .gitignore for tests, clang-tidy fixes * Label all ctest tests * ci : ctest uses -L main * Attempt at writing ctest_with_model * ci : get ci/run.sh working with test-model-load-cancel * ci : restrict .github/workflows/build.yml ctest to -L main * update requirements.txt * Disable test-model-load-cancel in make * Remove venv before creation * Restructure requirements.txt Top-level now imports the specific additional requirements for each python file. Using `pip install -r requirements.txt` will fail if versions become mismatched in the per-file requirements. * Make per-python-script requirements work alone This doesn't break the main requirements.txt. * Add comment * Add convert-persimmon-to-gguf.py to new requirements.txt scheme * Add check-requirements.sh script and GitHub workflow * Remove shellcheck installation step from workflow * Add nocleanup special arg * Fix merge see: https://github.com/ggerganov/llama.cpp/pull/4462#discussion_r1434593573 * reset to upstream/master * Redo changes for cancelling model load --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2023-12-22 06:19:36 +00:00
// Returns false if cancelled by progress_callback
static bool llm_load_tensors(
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama_model_loader & ml,
llama_model & model,
int n_gpu_layers,
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
enum llama_split_mode split_mode,
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
int main_gpu,
const float * tensor_split,
bool use_mlock,
llama_progress_callback progress_callback,
void * progress_callback_user_data) {
model.t_start_us = ggml_time_us();
auto & hparams = model.hparams;
ggml : group all experts in a single ggml_mul_mat_id (#6505) * ggml : group all experts in a single ggml_mul_mat_id cuda : improve mmid row copy * cuda : fix bin bcast with non-cont src0 * test-backend-ops : only run all mul mat tests for base types * llama : disable moe offloading with SYCL --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-18 13:18:48 +00:00
#ifdef GGML_USE_SYCL
// disable MoE with SYCL until mul_mat_id is updated
if (hparams.n_expert > 0) {
n_gpu_layers = 0;
}
#endif
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.split_mode = split_mode;
model.main_gpu = main_gpu;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
model.n_gpu_layers = n_gpu_layers;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
const int64_t n_layer = hparams.n_layer;
const int64_t i_gpu_start = std::max((int64_t) hparams.n_layer - n_gpu_layers, (int64_t) 0);
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
bool use_mmap_buffer = true;
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// there is very little benefit to offloading the input layer, so always keep it on the CPU
model.buft_input = llama_default_buffer_type_cpu(true);
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
//model.buft_input = llama_default_buffer_type_offload(main_gpu);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.buft_layer.resize(n_layer);
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// assign cpu layers
for (int64_t i = 0; i < i_gpu_start; ++i) {
model.buft_layer[i] = llama_default_buffer_type_cpu(true);
}
code : normalize enum names (#5697) * coda : normalize enum names ggml-ci * code : cont * code : cont
2024-02-25 10:09:09 +00:00
if (split_mode == LLAMA_SPLIT_MODE_LAYER) {
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// calculate the split points
Basic Vulkan Multi-GPU implementation (#5321) * Initial Vulkan multi-gpu implementation Move most global variables into backend context * Add names to backend device functions * Add further missing cleanup code * Reduce code duplication in tensor split layer assignment * generalize LLAMA_SPLIT_LAYER for all backends, do not expose device count and memory in llama.h * Only do device info print in the beginning and initialize one backend for cpu assist Add missing cleanup code * Rework backend memory management to make sure devices and buffers get properly allocated and freed * Rename cpu assist free function --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-07 06:54:50 +00:00
int device_count = llama_get_device_count();
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
bool all_zero = tensor_split == nullptr || std::all_of(tensor_split, tensor_split + device_count, [](float x) { return x == 0.0f; });
Basic Vulkan Multi-GPU implementation (#5321) * Initial Vulkan multi-gpu implementation Move most global variables into backend context * Add names to backend device functions * Add further missing cleanup code * Reduce code duplication in tensor split layer assignment * generalize LLAMA_SPLIT_LAYER for all backends, do not expose device count and memory in llama.h * Only do device info print in the beginning and initialize one backend for cpu assist Add missing cleanup code * Rework backend memory management to make sure devices and buffers get properly allocated and freed * Rename cpu assist free function --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-07 06:54:50 +00:00
std::vector<float> splits(device_count);
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
if (all_zero) {
// default split, by free memory
for (int i = 0; i < device_count; ++i) {
Basic Vulkan Multi-GPU implementation (#5321) * Initial Vulkan multi-gpu implementation Move most global variables into backend context * Add names to backend device functions * Add further missing cleanup code * Reduce code duplication in tensor split layer assignment * generalize LLAMA_SPLIT_LAYER for all backends, do not expose device count and memory in llama.h * Only do device info print in the beginning and initialize one backend for cpu assist Add missing cleanup code * Rework backend memory management to make sure devices and buffers get properly allocated and freed * Rename cpu assist free function --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-07 06:54:50 +00:00
splits[i] = llama_get_device_memory(i);
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
}
} else {
Basic Vulkan Multi-GPU implementation (#5321) * Initial Vulkan multi-gpu implementation Move most global variables into backend context * Add names to backend device functions * Add further missing cleanup code * Reduce code duplication in tensor split layer assignment * generalize LLAMA_SPLIT_LAYER for all backends, do not expose device count and memory in llama.h * Only do device info print in the beginning and initialize one backend for cpu assist Add missing cleanup code * Rework backend memory management to make sure devices and buffers get properly allocated and freed * Rename cpu assist free function --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-07 06:54:50 +00:00
std::copy(tensor_split, tensor_split + device_count, splits.begin());
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
}
// sum and normalize the splits to get the split points
float split_sum = 0.0f;
for (int i = 0; i < device_count; ++i) {
split_sum += splits[i];
splits[i] = split_sum;
}
for (int i = 0; i < device_count; ++i) {
splits[i] /= split_sum;
}
// assign the repeating layers to the devices according to the splits
int act_gpu_layers = std::min(n_gpu_layers, (int)n_layer + 1);
for (int64_t i = i_gpu_start; i < n_layer; ++i) {
Basic Vulkan Multi-GPU implementation (#5321) * Initial Vulkan multi-gpu implementation Move most global variables into backend context * Add names to backend device functions * Add further missing cleanup code * Reduce code duplication in tensor split layer assignment * generalize LLAMA_SPLIT_LAYER for all backends, do not expose device count and memory in llama.h * Only do device info print in the beginning and initialize one backend for cpu assist Add missing cleanup code * Rework backend memory management to make sure devices and buffers get properly allocated and freed * Rename cpu assist free function --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-07 06:54:50 +00:00
int layer_gpu = std::upper_bound(splits.begin(), splits.begin() + device_count, float(i - i_gpu_start)/act_gpu_layers) - splits.begin();
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.buft_layer[i] = llama_default_buffer_type_offload(layer_gpu);
}
// assign the output layer
if (n_gpu_layers > n_layer) {
Basic Vulkan Multi-GPU implementation (#5321) * Initial Vulkan multi-gpu implementation Move most global variables into backend context * Add names to backend device functions * Add further missing cleanup code * Reduce code duplication in tensor split layer assignment * generalize LLAMA_SPLIT_LAYER for all backends, do not expose device count and memory in llama.h * Only do device info print in the beginning and initialize one backend for cpu assist Add missing cleanup code * Rework backend memory management to make sure devices and buffers get properly allocated and freed * Rename cpu assist free function --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-07 06:54:50 +00:00
int layer_gpu = std::upper_bound(splits.begin(), splits.begin() + device_count, float(act_gpu_layers - 1)/act_gpu_layers) - splits.begin();
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.buft_output = llama_default_buffer_type_offload(layer_gpu);
} else {
model.buft_output = llama_default_buffer_type_cpu(true);
}
Basic Vulkan Multi-GPU implementation (#5321) * Initial Vulkan multi-gpu implementation Move most global variables into backend context * Add names to backend device functions * Add further missing cleanup code * Reduce code duplication in tensor split layer assignment * generalize LLAMA_SPLIT_LAYER for all backends, do not expose device count and memory in llama.h * Only do device info print in the beginning and initialize one backend for cpu assist Add missing cleanup code * Rework backend memory management to make sure devices and buffers get properly allocated and freed * Rename cpu assist free function --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-07 06:54:50 +00:00
} else {
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ggml_backend_buffer_type_t split_buft;
code : normalize enum names (#5697) * coda : normalize enum names ggml-ci * code : cont * code : cont
2024-02-25 10:09:09 +00:00
if (split_mode == LLAMA_SPLIT_MODE_ROW) {
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
split_buft = llama_default_buffer_type_split(main_gpu, tensor_split);
} else {
code : normalize enum names (#5697) * coda : normalize enum names ggml-ci * code : cont * code : cont
2024-02-25 10:09:09 +00:00
// LLAMA_SPLIT_MODE_NONE or LLAMA_SPLIT_MODE_LAYER in backends where it is not supported
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
split_buft = llama_default_buffer_type_offload(main_gpu);
}
// assign the repeating layers
for (int64_t i = i_gpu_start; i < n_layer; ++i) {
model.buft_layer[i] = {
split_buft,
llama_default_buffer_type_offload(main_gpu)
};
}
// assign the output layer
if (n_gpu_layers > n_layer) {
model.buft_output = {
split_buft,
llama_default_buffer_type_offload(main_gpu)
};
} else {
model.buft_output = llama_default_buffer_type_cpu(true);
}
}
// count used buffer types
std::map<ggml_backend_buffer_type_t, int> buft_layer_count;
buft_layer_count[model.buft_input.buft]++;
buft_layer_count[model.buft_input.buft_matrix]++;
buft_layer_count[model.buft_output.buft]++;
buft_layer_count[model.buft_output.buft_matrix]++;
for (int64_t i = 0; i < n_layer; ++i) {
buft_layer_count[model.buft_layer[i].buft]++;
buft_layer_count[model.buft_layer[i].buft_matrix]++;
}
// create one context per buffer type
llama : fix loading models with shared tok_embd and output (#5651) ggml-ci
2024-02-21 23:42:09 +00:00
size_t ctx_size = ggml_tensor_overhead()*(ml.n_tensors + 1); // +1 for models where tok_embd is duplicated as output
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
// for moe merged tensors
llama : fix compatibility with old 2 expert models (#6735)
2024-04-18 07:04:47 +00:00
ctx_size += ggml_tensor_overhead()*n_layer*3;
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
std::map<ggml_backend_buffer_type_t, ggml_context *> ctx_map;
for (auto & it : buft_layer_count) {
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
struct ggml_init_params params = {
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
/*.mem_size =*/ ctx_size,
/*.mem_buffer =*/ NULL,
/*.no_alloc =*/ true,
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
};
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ggml_context * ctx = ggml_init(params);
if (!ctx) {
throw std::runtime_error(format("failed to create context"));
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ctx_map[it.first] = ctx;
model.ctxs.push_back(ctx);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
LLAMA_LOG_INFO("%s: ggml ctx size = %7.2f MiB\n", __func__, model.ctxs.size()*ctx_size/1024.0/1024.0);
cuda : loading models directly into VRAM, norm calculation on GPU, broadcasting for ggml_mul (#1483) * Broadcasting for ggml_mul * CUDA kernel for ggml_mul, norms in VRAM * GPU weights not in RAM, direct loading with cuFile * fixup! GPU weights not in RAM, direct loading with cuFile * fixup! GPU weights not in RAM, direct loading with cuFile * define default model path once, sync path with readme (#1366) * ~7% faster Q5_1 AVX2 code (#1477) * convert.py: Support models which are stored in a single pytorch_model.bin (#1469) * Support models in a single pytorch_model.bin * Remove spurious line with typo * benchmark-matmul: Print the average of the test results (#1490) * Remove unused n_parts parameter (#1509) * Fixes #1511 lambda issue for w64devkit (mingw) (#1513) * Fix for w64devkit and mingw * make kv_f16 the default for api users (#1517) * minor : fix compile warnings * readme : adds WizardLM to the list of supported models (#1485) * main : make reverse prompt option act as a stop token in non-interactive mode (#1032) * Make reverse prompt option act as a stop token in non-interactive scenarios * Making requested review changes * Update gpt_params_parse and fix a merge error * Revert "Update gpt_params_parse and fix a merge error" This reverts commit 2bb2ff1748513591ad45b175a75ed1d8089d84c8. * Update gpt_params_parse and fix a merge error take 2 * examples : add persistent chat (#1495) * examples : add persistent chat * examples : fix whitespace --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * tests : add missing header * ggml : use F16 instead of F32 in Q4_0, Q4_1, Q8_0 (#1508) * ggml : use F16 instead of F32 in Q4_0, Q4_1 and Q8_0 * llama : bump LLAMA_FILE_VERSION to 3 * cuda : update Q4 and Q8 dequantize kernels * ggml : fix AVX dot products * readme : update performance table + hot topics * ggml : fix scalar implementation of Q4_1 dot * llama : fix compile warnings in llama_set_state_data() * llama : fix name shadowing and C4146 (#1526) * Fix name shadowing and C4146 * Fix if macros not using defined when required * Update llama-util.h Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Update llama-util.h Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Code style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix for mingw (#1462) * llama : add llama_init_backend() API (close #1527) * feature : add blis and other BLAS implementation support (#1502) * feature: add blis support * feature: allow all BLA_VENDOR to be assigned in cmake arguments. align with whisper.cpp pr 927 * fix: version detection for BLA_SIZEOF_INTEGER, recover min version of cmake * Fix typo in INTEGER Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Revert "feature : add blis and other BLAS implementation support (#1502)" This reverts commit 07e9ace0f9da424d82e75df969642522880feb92. * GPU weights not in RAM, direct loading with cuFile * llama : code style fixes + progress print fix * ggml : ggml_mul better broadcast support * cmake : workarounds for cufile when CMake version < 3.25 * gg rebase fixup * Loop in llama.cpp, fixed progress callback * Attempt clang-tidy fix * llama : fix vram size computation * Add forgotten fclose() --------- Co-authored-by: András Salamon <ott2@users.noreply.github.com> Co-authored-by: Ilya Kurdyukov <59548320+ilyakurdyukov@users.noreply.github.com> Co-authored-by: Tom Jobbins <784313+TheBloke@users.noreply.github.com> Co-authored-by: rankaiyx <rankaiyx@rankaiyx.com> Co-authored-by: Stephan Walter <stephan@walter.name> Co-authored-by: DannyDaemonic <DannyDaemonic@gmail.com> Co-authored-by: Erik Scholz <Green-Sky@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: David Kennedy <dakennedyd@gmail.com> Co-authored-by: Jason McCartney <jmac@theroot.org> Co-authored-by: Evan Jones <evan.q.jones@gmail.com> Co-authored-by: Maxime <672982+maximegmd@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Zenix <zenixls2@gmail.com>
2023-05-20 12:19:28 +00:00
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
// create tensors for the weights
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
{
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_embd = hparams.n_embd;
const int64_t n_embd_k_gqa = hparams.n_embd_k_gqa();
const int64_t n_embd_v_gqa = hparams.n_embd_v_gqa();
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
const int64_t n_embd_gqa = n_embd_v_gqa;
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_vocab = hparams.n_vocab;
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
const int64_t n_vocab_type = hparams.n_vocab_type;
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
const int64_t n_ff = hparams.n_ff;
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
const int64_t n_expert = hparams.n_expert;
if (n_expert > 0 && hparams.n_expert_used == 0) {
throw std::runtime_error("model has expert layers but no expert layers are used");
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
GGML_ASSERT(n_embd_gqa == n_embd_k_gqa);
ggml_context * ctx_input = ctx_map.at(model.buft_input.buft);
ggml_context * ctx_output = ctx_map.at(model.buft_output.buft);
ggml_context * ctx_output_split = ctx_map.at(model.buft_output.buft_matrix);
auto ctx_for_layer = [&](int i) { return ctx_map.at(model.buft_layer[i].buft); };
auto ctx_for_layer_split = [&](int i) { return ctx_map.at(model.buft_layer[i].buft_matrix); };
model.layers.resize(n_layer);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
const auto tn = LLM_TN(model.arch);
switch (model.arch) {
case LLM_ARCH_LLAMA:
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
case LLM_ARCH_REFACT:
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
case LLM_ARCH_MINICPM:
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// output
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
if (model.arch != LLM_ARCH_MINICPM){
llama : assume tied weights if lm_head/output weights is missing (#5824) This is to support model configurations with "tie_word_embeddings" set to true. Co-authored-by: Don Mahurin <2797413+dmahurin@users.noreply.github.com>
2024-03-08 10:41:50 +00:00
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, false);
// if output is NULL, init from the input tok embed
if (model.output == NULL) {
model.output = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
ml.n_created--; // artificial tensor
ml.size_data += ggml_nbytes(model.output);
}
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
cuda : loading models directly into VRAM, norm calculation on GPU, broadcasting for ggml_mul (#1483) * Broadcasting for ggml_mul * CUDA kernel for ggml_mul, norms in VRAM * GPU weights not in RAM, direct loading with cuFile * fixup! GPU weights not in RAM, direct loading with cuFile * fixup! GPU weights not in RAM, direct loading with cuFile * define default model path once, sync path with readme (#1366) * ~7% faster Q5_1 AVX2 code (#1477) * convert.py: Support models which are stored in a single pytorch_model.bin (#1469) * Support models in a single pytorch_model.bin * Remove spurious line with typo * benchmark-matmul: Print the average of the test results (#1490) * Remove unused n_parts parameter (#1509) * Fixes #1511 lambda issue for w64devkit (mingw) (#1513) * Fix for w64devkit and mingw * make kv_f16 the default for api users (#1517) * minor : fix compile warnings * readme : adds WizardLM to the list of supported models (#1485) * main : make reverse prompt option act as a stop token in non-interactive mode (#1032) * Make reverse prompt option act as a stop token in non-interactive scenarios * Making requested review changes * Update gpt_params_parse and fix a merge error * Revert "Update gpt_params_parse and fix a merge error" This reverts commit 2bb2ff1748513591ad45b175a75ed1d8089d84c8. * Update gpt_params_parse and fix a merge error take 2 * examples : add persistent chat (#1495) * examples : add persistent chat * examples : fix whitespace --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * tests : add missing header * ggml : use F16 instead of F32 in Q4_0, Q4_1, Q8_0 (#1508) * ggml : use F16 instead of F32 in Q4_0, Q4_1 and Q8_0 * llama : bump LLAMA_FILE_VERSION to 3 * cuda : update Q4 and Q8 dequantize kernels * ggml : fix AVX dot products * readme : update performance table + hot topics * ggml : fix scalar implementation of Q4_1 dot * llama : fix compile warnings in llama_set_state_data() * llama : fix name shadowing and C4146 (#1526) * Fix name shadowing and C4146 * Fix if macros not using defined when required * Update llama-util.h Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Update llama-util.h Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Code style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix for mingw (#1462) * llama : add llama_init_backend() API (close #1527) * feature : add blis and other BLAS implementation support (#1502) * feature: add blis support * feature: allow all BLA_VENDOR to be assigned in cmake arguments. align with whisper.cpp pr 927 * fix: version detection for BLA_SIZEOF_INTEGER, recover min version of cmake * Fix typo in INTEGER Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Revert "feature : add blis and other BLAS implementation support (#1502)" This reverts commit 07e9ace0f9da424d82e75df969642522880feb92. * GPU weights not in RAM, direct loading with cuFile * llama : code style fixes + progress print fix * ggml : ggml_mul better broadcast support * cmake : workarounds for cufile when CMake version < 3.25 * gg rebase fixup * Loop in llama.cpp, fixed progress callback * Attempt clang-tidy fix * llama : fix vram size computation * Add forgotten fclose() --------- Co-authored-by: András Salamon <ott2@users.noreply.github.com> Co-authored-by: Ilya Kurdyukov <59548320+ilyakurdyukov@users.noreply.github.com> Co-authored-by: Tom Jobbins <784313+TheBloke@users.noreply.github.com> Co-authored-by: rankaiyx <rankaiyx@rankaiyx.com> Co-authored-by: Stephan Walter <stephan@walter.name> Co-authored-by: DannyDaemonic <DannyDaemonic@gmail.com> Co-authored-by: Erik Scholz <Green-Sky@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: David Kennedy <dakennedyd@gmail.com> Co-authored-by: Jason McCartney <jmac@theroot.org> Co-authored-by: Evan Jones <evan.q.jones@gmail.com> Co-authored-by: Maxime <672982+maximegmd@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Zenix <zenixls2@gmail.com>
2023-05-20 12:19:28 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
auto & layer = model.layers[i];
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd});
layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa});
layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llama : support optional tensors (#4283)
2023-12-01 18:35:03 +00:00
// optional bias tensors
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.bq = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q, "bias", i), {n_embd}, false);
layer.bk = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K, "bias", i), {n_embd_gqa}, false);
layer.bv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_V, "bias", i), {n_embd_gqa}, false);
layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, false);
llama : support attention bias on LLaMA architecture (#4283) * Support attention_bias on LLaMA architecture QKVO bias, should fix InternLM (https://github.com/ggerganov/llama.cpp/issues/3133) and works for LLaMAfied Qwen models (https://github.com/ggerganov/llama.cpp/pull/3743#issuecomment-1825923608). * check existence of qkvo bias while loading llama models Tested on LLaMA2, CUDA and CPU. * Update llama.cpp
2023-12-01 18:17:06 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
if (n_expert == 0) {
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff});
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd});
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
} else {
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
layer.ffn_gate_inp = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert});
layer.ffn_gate_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff, n_expert}, false);
if (layer.ffn_gate_exps) {
layer.ffn_down_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), { n_ff, n_embd, n_expert});
layer.ffn_up_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP_EXPS, "weight", i), {n_embd, n_ff, n_expert});
} else {
// merge split expert into a single tensor for compatibility with older models
// requires disabling mmap
use_mmap_buffer = false;
ggml_type type_gate = ml.require_tensor_meta(tn(LLM_TENSOR_FFN_GATE_EXP, "weight", i, 0).c_str())->type;
ggml_type type_down = ml.require_tensor_meta(tn(LLM_TENSOR_FFN_DOWN_EXP, "weight", i, 0).c_str())->type;
ggml_type type_up = ml.require_tensor_meta(tn(LLM_TENSOR_FFN_UP_EXP, "weight", i, 0).c_str())->type;
layer.ffn_gate_exps = ggml_new_tensor_3d(ctx_split, type_gate, n_embd, n_ff, n_expert);
layer.ffn_down_exps = ggml_new_tensor_3d(ctx_split, type_down, n_ff, n_embd, n_expert);
layer.ffn_up_exps = ggml_new_tensor_3d(ctx_split, type_up, n_embd, n_ff, n_expert);
ggml_set_name(layer.ffn_gate_exps, tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i).c_str());
ggml_set_name(layer.ffn_down_exps, tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i).c_str());
ggml_set_name(layer.ffn_up_exps, tn(LLM_TENSOR_FFN_UP_EXPS, "weight", i).c_str());
for (uint32_t x = 0; x < n_expert; ++x) {
// the individual experts are loaded into a view of the merged tensor
ml.create_tensor_as_view(ctx_split, layer.ffn_gate_exps, tn(LLM_TENSOR_FFN_GATE_EXP, "weight", i, x), { n_embd, n_ff }, layer.ffn_gate_exps->nb[2]*x);
ml.create_tensor_as_view(ctx_split, layer.ffn_down_exps, tn(LLM_TENSOR_FFN_DOWN_EXP, "weight", i, x), { n_ff, n_embd }, layer.ffn_down_exps->nb[2]*x);
ml.create_tensor_as_view(ctx_split, layer.ffn_up_exps, tn(LLM_TENSOR_FFN_UP_EXP, "weight", i, x), { n_embd, n_ff }, layer.ffn_up_exps->nb[2]*x);
}
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
}
}
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
}
} break;
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
case LLM_ARCH_GROK:
{
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
if (n_expert == 0) {
throw std::runtime_error("Grok model cannot have zero experts");
}
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
// output
{
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, false);
// if output is NULL, init from the input tok embed
if (model.output == NULL) {
model.output = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
ml.n_created--; // artificial tensor
ml.size_data += ggml_nbytes(model.output);
}
}
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
auto & layer = model.layers[i];
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd});
layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa});
layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
layer.attn_out_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd});
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
layer.ffn_gate_inp = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert});
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
layer.ffn_gate_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff, n_expert}, false);
if (layer.ffn_gate_exps) {
layer.ffn_down_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), { n_ff, n_embd, n_expert});
layer.ffn_up_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP_EXPS, "weight", i), {n_embd, n_ff, n_expert});
} else {
// merge split expert into a single tensor for compatibility with older models
// requires disabling mmap
use_mmap_buffer = false;
ggml_type type_gate = ml.require_tensor_meta(tn(LLM_TENSOR_FFN_GATE_EXP, "weight", i, 0).c_str())->type;
ggml_type type_down = ml.require_tensor_meta(tn(LLM_TENSOR_FFN_DOWN_EXP, "weight", i, 0).c_str())->type;
ggml_type type_up = ml.require_tensor_meta(tn(LLM_TENSOR_FFN_UP_EXP, "weight", i, 0).c_str())->type;
layer.ffn_gate_exps = ggml_new_tensor_3d(ctx_split, type_gate, n_embd, n_ff, n_expert);
layer.ffn_down_exps = ggml_new_tensor_3d(ctx_split, type_down, n_ff, n_embd, n_expert);
layer.ffn_up_exps = ggml_new_tensor_3d(ctx_split, type_up, n_embd, n_ff, n_expert);
ggml_set_name(layer.ffn_gate_exps, tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i).c_str());
ggml_set_name(layer.ffn_down_exps, tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i).c_str());
ggml_set_name(layer.ffn_up_exps, tn(LLM_TENSOR_FFN_UP_EXPS, "weight", i).c_str());
for (uint32_t x = 0; x < n_expert; ++x) {
// the individual experts are loaded into a view of the merged tensor
ml.create_tensor_as_view(ctx_split, layer.ffn_gate_exps, tn(LLM_TENSOR_FFN_GATE_EXP, "weight", i, x), { n_embd, n_ff }, layer.ffn_gate_exps->nb[2]*x);
ml.create_tensor_as_view(ctx_split, layer.ffn_down_exps, tn(LLM_TENSOR_FFN_DOWN_EXP, "weight", i, x), { n_ff, n_embd }, layer.ffn_down_exps->nb[2]*x);
ml.create_tensor_as_view(ctx_split, layer.ffn_up_exps, tn(LLM_TENSOR_FFN_UP_EXP, "weight", i, x), { n_embd, n_ff }, layer.ffn_up_exps->nb[2]*x);
}
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
}
layer.layer_out_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_LAYER_OUT_NORM, "weight", i), {n_embd});
}
} break;
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
case LLM_ARCH_DBRX:
{
if (n_expert == 0) {
throw std::runtime_error("DBRX model cannot have zero experts");
}
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
// output
{
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab});
}
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
auto & layer = model.layers[i];
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
layer.attn_out_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd});
layer.ffn_gate_inp = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert});
layer.ffn_gate_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff, n_expert});
layer.ffn_down_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff, n_embd, n_expert});
layer.ffn_up_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP_EXPS, "weight", i), {n_embd, n_ff, n_expert});
}
} break;
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
case LLM_ARCH_BAICHUAN:
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab});
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
auto & layer = model.layers[i];
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd});
layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa});
layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
cuda : loading models directly into VRAM, norm calculation on GPU, broadcasting for ggml_mul (#1483) * Broadcasting for ggml_mul * CUDA kernel for ggml_mul, norms in VRAM * GPU weights not in RAM, direct loading with cuFile * fixup! GPU weights not in RAM, direct loading with cuFile * fixup! GPU weights not in RAM, direct loading with cuFile * define default model path once, sync path with readme (#1366) * ~7% faster Q5_1 AVX2 code (#1477) * convert.py: Support models which are stored in a single pytorch_model.bin (#1469) * Support models in a single pytorch_model.bin * Remove spurious line with typo * benchmark-matmul: Print the average of the test results (#1490) * Remove unused n_parts parameter (#1509) * Fixes #1511 lambda issue for w64devkit (mingw) (#1513) * Fix for w64devkit and mingw * make kv_f16 the default for api users (#1517) * minor : fix compile warnings * readme : adds WizardLM to the list of supported models (#1485) * main : make reverse prompt option act as a stop token in non-interactive mode (#1032) * Make reverse prompt option act as a stop token in non-interactive scenarios * Making requested review changes * Update gpt_params_parse and fix a merge error * Revert "Update gpt_params_parse and fix a merge error" This reverts commit 2bb2ff1748513591ad45b175a75ed1d8089d84c8. * Update gpt_params_parse and fix a merge error take 2 * examples : add persistent chat (#1495) * examples : add persistent chat * examples : fix whitespace --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * tests : add missing header * ggml : use F16 instead of F32 in Q4_0, Q4_1, Q8_0 (#1508) * ggml : use F16 instead of F32 in Q4_0, Q4_1 and Q8_0 * llama : bump LLAMA_FILE_VERSION to 3 * cuda : update Q4 and Q8 dequantize kernels * ggml : fix AVX dot products * readme : update performance table + hot topics * ggml : fix scalar implementation of Q4_1 dot * llama : fix compile warnings in llama_set_state_data() * llama : fix name shadowing and C4146 (#1526) * Fix name shadowing and C4146 * Fix if macros not using defined when required * Update llama-util.h Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Update llama-util.h Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Code style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix for mingw (#1462) * llama : add llama_init_backend() API (close #1527) * feature : add blis and other BLAS implementation support (#1502) * feature: add blis support * feature: allow all BLA_VENDOR to be assigned in cmake arguments. align with whisper.cpp pr 927 * fix: version detection for BLA_SIZEOF_INTEGER, recover min version of cmake * Fix typo in INTEGER Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Revert "feature : add blis and other BLAS implementation support (#1502)" This reverts commit 07e9ace0f9da424d82e75df969642522880feb92. * GPU weights not in RAM, direct loading with cuFile * llama : code style fixes + progress print fix * ggml : ggml_mul better broadcast support * cmake : workarounds for cufile when CMake version < 3.25 * gg rebase fixup * Loop in llama.cpp, fixed progress callback * Attempt clang-tidy fix * llama : fix vram size computation * Add forgotten fclose() --------- Co-authored-by: András Salamon <ott2@users.noreply.github.com> Co-authored-by: Ilya Kurdyukov <59548320+ilyakurdyukov@users.noreply.github.com> Co-authored-by: Tom Jobbins <784313+TheBloke@users.noreply.github.com> Co-authored-by: rankaiyx <rankaiyx@rankaiyx.com> Co-authored-by: Stephan Walter <stephan@walter.name> Co-authored-by: DannyDaemonic <DannyDaemonic@gmail.com> Co-authored-by: Erik Scholz <Green-Sky@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: David Kennedy <dakennedyd@gmail.com> Co-authored-by: Jason McCartney <jmac@theroot.org> Co-authored-by: Evan Jones <evan.q.jones@gmail.com> Co-authored-by: Maxime <672982+maximegmd@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Zenix <zenixls2@gmail.com>
2023-05-20 12:19:28 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff});
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd});
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
} break;
case LLM_ARCH_FALCON:
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
// output
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd});
mpt : implement backwards compatiblity with duped output tensor (#6139)
2024-03-18 16:49:02 +00:00
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, false);
if (!model.output) {
llama : fix falcon arch for tied output embeddings (#4978) * falcon arch fix for tied output embeddings * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-18 22:12:15 +00:00
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); // needs to be on GPU
ml.n_created--; // artificial tensor
llama : fix loading models with shared tok_embd and output (#5651) ggml-ci
2024-02-21 23:42:09 +00:00
ml.size_data += ggml_nbytes(model.output);
llama : fix falcon arch for tied output embeddings (#4978) * falcon arch fix for tied output embeddings * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-18 22:12:15 +00:00
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
auto & layer = model.layers[i];
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.attn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd});
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
layer.attn_norm_2 = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM_2, "weight", i), {n_embd}, false);
layer.attn_norm_2_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM_2, "bias", i), {n_embd}, false);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd});
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
} break;
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
case LLM_ARCH_STARCODER:
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
model.pos_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_POS_EMBD, "weight"), {n_embd, hparams.n_ctx_train});
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
// output
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab});
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
auto & layer = model.layers[i];
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.attn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd});
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa});
layer.bqkv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd + 2*n_embd_gqa});
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd});
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
layer.ffn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd});
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd});
layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd});
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
layer.ffn_up_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff});
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
}
} break;
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
case LLM_ARCH_PERSIMMON:
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab});
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
auto & layer = model.layers[i];
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.attn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd});
layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa});
layer.bqkv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd + 2*n_embd_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd});
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd});
layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd});
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
layer.ffn_up_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff});
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
layer.ffn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd});
layer.attn_q_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {64});
layer.attn_q_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_NORM, "bias", i), {64});
layer.attn_k_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {64});
layer.attn_k_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K_NORM, "bias", i), {64});
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
}
} break;
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
case LLM_ARCH_BERT:
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
case LLM_ARCH_NOMIC_BERT:
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
{
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
model.type_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_TYPES, "weight"), {n_embd, n_vocab_type});
if (model.arch == LLM_ARCH_BERT) {
model.pos_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_POS_EMBD, "weight"), {n_embd, hparams.n_ctx_train});
}
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
model.tok_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight"), {n_embd});
model.tok_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD_NORM, "bias"), {n_embd});
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
auto & layer = model.layers[i];
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
if (model.arch == LLM_ARCH_BERT) {
layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd});
layer.bq = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q, "bias", i), {n_embd});
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa});
layer.bk = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K, "bias", i), {n_embd_gqa});
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa});
layer.bv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_V, "bias", i), {n_embd_gqa});
} else {
layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa});
}
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
layer.attn_out_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd});
layer.attn_out_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT_NORM, "bias", i), {n_embd});
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd});
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
if (model.arch == LLM_ARCH_BERT) {
layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd});
layer.ffn_up_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff});
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd});
} else {
layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff});
}
layer.layer_out_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_LAYER_OUT_NORM, "weight", i), {n_embd});
layer.layer_out_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_LAYER_OUT_NORM, "bias", i), {n_embd});
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
}
} break;
case LLM_ARCH_BLOOM:
{
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
model.tok_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight"), {n_embd});
model.tok_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD_NORM, "bias"), {n_embd});
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
// output
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab});
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
auto & layer = model.layers[i];
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.attn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd});
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa});
layer.bqkv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd + 2*n_embd_gqa});
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd});
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
layer.ffn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd});
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd});
layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd});
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
layer.ffn_up_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff});
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
}
} break;
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
case LLM_ARCH_MPT:
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
llama : add SEA-LION support (#6448) * initial commit for sealion support * add sealion support * minor fix * q/k ln and pos_embd only if required * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * minor : clear whitespaces --------- Co-authored-by: bryan <bryansiow@aisingapore.org> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 18:05:10 +00:00
model.pos_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_POS_EMBD, "weight"), {n_embd, hparams.n_ctx_train}, false);
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
// output
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
mpt : add optional bias tensors (#5638) Update for MPT with optional bias parameters: to work with PhoGPT and SEA-LION models that were pre-trained with 'bias'.
2024-02-22 08:15:13 +00:00
model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd}, false);
mpt : implement backwards compatiblity with duped output tensor (#6139)
2024-03-18 16:49:02 +00:00
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, false);
if (!model.output) {
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); // needs to be on GPU
ml.n_created--; // artificial tensor
ml.size_data += ggml_nbytes(model.output);
}
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
auto & layer = model.layers[i];
mpt : add optional bias tensors (#5638) Update for MPT with optional bias parameters: to work with PhoGPT and SEA-LION models that were pre-trained with 'bias'.
2024-02-22 08:15:13 +00:00
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.attn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd}, false);
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa});
mpt : add optional bias tensors (#5638) Update for MPT with optional bias parameters: to work with PhoGPT and SEA-LION models that were pre-trained with 'bias'.
2024-02-22 08:15:13 +00:00
layer.bqkv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd + 2*n_embd_gqa}, false);
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
mpt : add optional bias tensors (#5638) Update for MPT with optional bias parameters: to work with PhoGPT and SEA-LION models that were pre-trained with 'bias'.
2024-02-22 08:15:13 +00:00
layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}, false);
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
mpt : add optional bias tensors (#5638) Update for MPT with optional bias parameters: to work with PhoGPT and SEA-LION models that were pre-trained with 'bias'.
2024-02-22 08:15:13 +00:00
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
layer.ffn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd}, false);
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd});
layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}, false);
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
layer.ffn_up_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff}, false);
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
llama : add SEA-LION support (#6448) * initial commit for sealion support * add sealion support * minor fix * q/k ln and pos_embd only if required * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * minor : clear whitespaces --------- Co-authored-by: bryan <bryansiow@aisingapore.org> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 18:05:10 +00:00
layer.attn_q_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd}, false);
layer.attn_q_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_NORM, "bias", i), {n_embd}, false);
layer.attn_k_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd}, false);
layer.attn_k_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K_NORM, "bias", i), {n_embd}, false);
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
// AWQ ScaleActivation layer
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_act = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_ACT, "scales", i), {n_ff}, false);
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
}
} break;
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
case LLM_ARCH_STABLELM:
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
// output
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd});
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab});
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
auto & layer = model.layers[i];
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.attn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd});
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd});
layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa});
layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
llama : support StableLM 2 1.6B (#5052) * llama : support StableLM 2 1.6B * convert : fix Qwen's set_vocab wrongly naming all special tokens [PAD{id}] * convert : refactor Qwen's set_vocab to use it for StableLM 2 too * nix : add tiktoken to llama-python-extra * convert : use presence of tokenizer.json to determine StableLM tokenizer loader It's a less arbitrary heuristic than the vocab size.
2024-01-22 11:21:52 +00:00
// optional bias tensors, present in Stable LM 2 1.6B
layer.bq = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q, "bias", i), {n_embd}, false);
layer.bk = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K, "bias", i), {n_embd_gqa}, false);
layer.bv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_V, "bias", i), {n_embd_gqa}, false);
llama : add StableLM2 12B (#6635) * StableLM2 12B support for huggingface -> GGUF * StableLM12 tensormapping and constants * StableLM-2-12b model support * fix * Added 12B support * Removed autoformatting; resolved bug where model_arch was not selecting StableLM2 * Formatting * Do QK norm stacking in model conversion step * Converge StableLM and StableLM2 code to simplify graph construction * Fix accidental removal * Removed warnings * Revert formatter * Move QK norm stack to private function so it's easier to read * refactor stablelm graph builder to support 1.6, 3b and 12b more efficiently * Proper check for None type for new_name to avoid crash; formatting; revert change to base class `write_tensors()` * Format * Formatting * format Co-authored-by: compilade <git@compilade.net> * Fix incorrect check for K norm * space after commas; Keep indentation multiple of 4 spaces * Flake8 format * Removed unnecessary conditional branches * Removed unused comment * Fixed incorrect tensor passing * Format --------- Co-authored-by: compilade <git@compilade.net>
2024-04-16 15:48:35 +00:00
// optional q and k layernorms, present in StableLM 2 12B
layer.attn_q_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {hparams.n_embd_head_k, hparams.n_head}, false);
layer.attn_k_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {hparams.n_embd_head_k, hparams.n_head_kv}, false);
// optional FFN norm, not present in StableLM 2 12B which uses parallel residual
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, false);
layer.ffn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd}, false);
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff});
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd});
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
}
} break;
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
case LLM_ARCH_QWEN:
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// output
{
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab});
}
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
auto & layer = model.layers[i];
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd*3});
layer.bqkv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd*3});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff/2});
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff/2, n_embd});
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff/2});
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
}
} break;
llama : support upcoming Qwen2 (#5037)
2024-01-19 11:53:13 +00:00
case LLM_ARCH_QWEN2:
{
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
// output
{
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
Qwen2 : assume tied weights if lm_head/output weights is missing (#6738)
2024-04-18 11:38:04 +00:00
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, false);
// if output is NULL, init from the input tok embed
if (model.output == NULL) {
model.output = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
ml.n_created--; // artificial tensor
ml.size_data += ggml_nbytes(model.output);
}
llama : support upcoming Qwen2 (#5037)
2024-01-19 11:53:13 +00:00
}
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
auto & layer = model.layers[i];
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd});
layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa});
layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
// optional bias tensors
layer.bq = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q, "bias", i), {n_embd});
layer.bk = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K, "bias", i), {n_embd_gqa});
layer.bv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_V, "bias", i), {n_embd_gqa});
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff});
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd});
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
}
} break;
llama : add qwen2moe (#6074) * support qwen2moe * fix-review * metal : support unary ops for nelements % 4 != 0 * metal : require contiguousness for float4 unary kernels * metal : require contiguousness for float4 unary kernels (cont) * fix-review * names : for brevity "SHARED_EXP" -> "SHEXP" * llama : reuse build_moe_ffn() * llama : add model type name --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-16 15:40:48 +00:00
case LLM_ARCH_QWEN2MOE:
{
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
// output
{
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab});
}
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
auto & layer = model.layers[i];
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd});
layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa});
layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
// optional bias tensors
layer.bq = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q, "bias", i), {n_embd});
layer.bk = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K, "bias", i), {n_embd_gqa});
layer.bv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_V, "bias", i), {n_embd_gqa});
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
layer.ffn_gate_inp = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert});
GGML_ASSERT(hparams.n_expert > 0);
GGML_ASSERT(hparams.n_expert_used > 0);
// MoE branch
auto n_ff_exp = n_ff / hparams.n_expert_used;
layer.ffn_gate_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), { n_embd, n_ff_exp, n_expert});
layer.ffn_down_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp, n_embd, n_expert});
layer.ffn_up_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP_EXPS, "weight", i), { n_embd, n_ff_exp, n_expert});
// Shared expert branch
layer.ffn_gate_inp_shexp = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_GATE_INP_SHEXP, "weight", i), {n_embd});
layer.ffn_gate_shexp = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff});
layer.ffn_down_shexp = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), { n_ff, n_embd});
layer.ffn_up_shexp = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP_SHEXP, "weight", i), {n_embd, n_ff});
}
} break;
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
case LLM_ARCH_PHI2:
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
// output
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab});
model.output_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT, "bias"), {n_vocab});
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
auto & layer = model.layers[i];
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.attn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd});
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
convert : update phi-2 to latest HF repo (#4903) * convert : update phi-2 to latest HF repo ggml-ci * py : try to fix flake stuff
2024-01-13 11:44:37 +00:00
layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}, false);
layer.bqkv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd + 2*n_embd_gqa}, false);
if (layer.wqkv == nullptr) {
layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd});
layer.bq = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q, "bias", i), {n_embd});
layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa});
layer.bk = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K, "bias", i), {n_embd_gqa});
layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa});
layer.bv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_V, "bias", i), {n_embd_gqa});
}
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd});
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd});
layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd});
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
layer.ffn_up_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff});
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
}
} break;
llama : add phi3 support (#6852) * add explicit phi3 support * add explicit phi3 support * remove unused code * convert : add BOS token * llama : match EOT token <|end|> * llama : minor / style * llama : tabs -> spaces * convert : fix lint checks --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-24 07:00:37 +00:00
case LLM_ARCH_PHI3:
{
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab });
// output
{
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), { n_embd });
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), { n_embd, n_vocab });
}
for (int i = 0; i < n_layer; ++i) {
ggml_context* ctx_layer = ctx_for_layer(i);
ggml_context* ctx_split = ctx_for_layer_split(i);
auto& layer = model.layers[i];
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), { n_embd });
layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), { n_embd, n_embd + 2 * n_embd_gqa }, false);
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_embd, n_embd });
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), { n_embd });
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd });
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), { n_embd, 2 * n_ff });
}
} break;
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
case LLM_ARCH_PLAMO:
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
// output
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab});
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
auto & layer = model.layers[i];
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd});
layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa});
layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff});
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd});
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
}
} break;
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
case LLM_ARCH_GPT2:
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
model.pos_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_POS_EMBD, "weight"), {n_embd, hparams.n_ctx_train});
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
// output
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab});
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
auto & layer = model.layers[i];
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.attn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd});
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa});
layer.bqkv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd + 2*n_embd_gqa});
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd});
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
layer.ffn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd});
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd});
layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd});
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
layer.ffn_up_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff});
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
}
} break;
llama : add CodeShell support (#5016) * llama: add codeshell support * llama.cpp: fix codeshell with NeoX rope Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-19 09:07:27 +00:00
case LLM_ARCH_CODESHELL:
{
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
// output
{
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab});
}
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
auto & layer = model.layers[i];
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.attn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd});
layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa});
layer.bqkv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd + 2*n_embd_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd});
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
layer.ffn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd});
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd});
layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd});
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
layer.ffn_up_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff});
}
} break;
llama : add support for Orion-14B (#5118) * add support for Orion-14B(https://huggingface.co/OrionStarAI/Orion-14B-Chat) * flake8 support * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * Update llama.cpp * Update llama.cpp --------- Co-authored-by: lixiaopu <lixiaopu@cmcm.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-01-28 08:00:30 +00:00
case LLM_ARCH_ORION:
{
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
{
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab});
}
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
auto & layer = model.layers[i];
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.attn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd});
layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd});
layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa});
layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
layer.ffn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd});
layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff});
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd});
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
}
} break;
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
case LLM_ARCH_INTERNLM2:
{
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
// output
{
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab});
}
llama : add support for Orion-14B (#5118) * add support for Orion-14B(https://huggingface.co/OrionStarAI/Orion-14B-Chat) * flake8 support * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * Update llama.cpp * Update llama.cpp --------- Co-authored-by: lixiaopu <lixiaopu@cmcm.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-01-28 08:00:30 +00:00
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
auto & layer = model.layers[i];
llama : add support for Orion-14B (#5118) * add support for Orion-14B(https://huggingface.co/OrionStarAI/Orion-14B-Chat) * flake8 support * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * Update llama.cpp * Update llama.cpp --------- Co-authored-by: lixiaopu <lixiaopu@cmcm.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-01-28 08:00:30 +00:00
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
// layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa});
layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd});
layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa});
layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff});
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd});
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
}
} break;
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
case LLM_ARCH_GEMMA:
{
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
// output
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
gemma : allow offloading the output tensor (#5646)
2024-02-21 21:18:23 +00:00
model.output = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); // same as tok_embd, duplicated to allow offloading
ml.n_created--; // artificial tensor
llama : fix loading models with shared tok_embd and output (#5651) ggml-ci
2024-02-21 23:42:09 +00:00
ml.size_data += ggml_nbytes(model.output);
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
const int64_t n_ff = hparams.n_ff;
const int64_t n_embd_head_k = hparams.n_embd_head_k;
const int64_t n_embd_k_gqa = hparams.n_embd_k_gqa();
const int64_t n_embd_v_gqa = hparams.n_embd_v_gqa();
for (uint32_t i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
auto & layer = model.layers[i];
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd_head_k * hparams.n_head});
layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_k_gqa});
layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_v_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * hparams.n_head, n_embd});
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff});
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd});
}
} break;
llama : add StarCoder2 support (#5795) * Add support for starcoder2 * handle rope type * skip rope freq and rotary embeddings from being serialized * resolve comments * Update llama.cpp * remove redundant changes * handle `rope-theta` * llama : change starcoder2 rope type * address comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-01 19:30:46 +00:00
case LLM_ARCH_STARCODER2:
{
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
// output
{
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, false);
// if output is NULL, init from the input tok embed
if (model.output == NULL) {
model.output = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
ml.n_created--; // artificial tensor
ml.size_data += ggml_nbytes(model.output);
}
}
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
auto & layer = model.layers[i];
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.attn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd});
layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd});
layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa});
layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
// optional bias tensors
layer.bq = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q, "bias", i), {n_embd});
layer.bk = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K, "bias", i), {n_embd_gqa});
layer.bv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_V, "bias", i), {n_embd_gqa});
layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd});
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
layer.ffn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd});
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd});
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
// optional bias tensors
layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd});
layer.ffn_up_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP , "bias", i), { n_ff});
}
} break;
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
case LLM_ARCH_MAMBA:
{
const int64_t d_conv = hparams.ssm_d_conv;
const int64_t d_inner = hparams.ssm_d_inner;
const int64_t d_state = hparams.ssm_d_state;
const int64_t dt_rank = hparams.ssm_dt_rank;
// only an expansion factor of 2 is supported for now
GGML_ASSERT(2 * n_embd == d_inner);
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
// output
{
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, false);
// if output is NULL, init from the input tok embed, duplicated to allow offloading
if (model.output == NULL) {
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
ml.n_created--; // artificial tensor
ml.size_data += ggml_nbytes(model.output);
}
}
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
auto & layer = model.layers[i];
// norm
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.ssm_in = ml.create_tensor(ctx_split, tn(LLM_TENSOR_SSM_IN, "weight", i), {n_embd, 2*d_inner});
layer.ssm_conv1d = ml.create_tensor(ctx_split, tn(LLM_TENSOR_SSM_CONV1D, "weight", i), {d_conv, d_inner});
layer.ssm_conv1d_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_SSM_CONV1D, "bias", i), {d_inner});
layer.ssm_x = ml.create_tensor(ctx_split, tn(LLM_TENSOR_SSM_X, "weight", i), {d_inner, dt_rank + 2*d_state});
layer.ssm_dt = ml.create_tensor(ctx_split, tn(LLM_TENSOR_SSM_DT, "weight", i), {dt_rank, d_inner});
layer.ssm_dt_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_SSM_DT, "bias", i), {d_inner});
// no "weight" suffix for these
layer.ssm_a = ml.create_tensor(ctx_split, tn(LLM_TENSOR_SSM_A, i), {d_state, d_inner});
layer.ssm_d = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_SSM_D, i), {d_inner});
// out_proj
layer.ssm_out = ml.create_tensor(ctx_split, tn(LLM_TENSOR_SSM_OUT, "weight", i), {d_inner, n_embd});
}
} break;
[Model] Add support for xverse (#6301) * Support xverse model convert to gguf format. * 1. Convert xverse models to gguf; 2. Add LLM_ARCH_XVERSE inference in llama.cpp; 3. Add xverse item in Supported models in README.md; * * gguf-py: remove redundant logs * llama: remove the init_mapping_prefetch custom parameter * llama.cpp: Include the changes from #6122 to exclude the unused outputs of the last layers. * - Fix format issues - Remove duplicate set kqv_out to llm_build_kv * Update llama.cpp --------- Co-authored-by: willhe <willhe@xverse.cn> Co-authored-by: willhe <hexin@xverse.cn>
2024-03-29 13:37:03 +00:00
case LLM_ARCH_XVERSE:
{
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
{
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab});
}
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
auto & layer = model.layers[i];
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd});
layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa});
layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd});
layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff});
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd});
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
}
} break;
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
case LLM_ARCH_COMMAND_R:
{
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
// output
{
model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd});
// init output from the input tok embed
model.output = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
ml.n_created--; // artificial tensor
ml.size_data += ggml_nbytes(model.output);
}
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_layer = ctx_for_layer(i);
ggml_context * ctx_split = ctx_for_layer_split(i);
auto & layer = model.layers[i];
layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd});
llama : add Command R Plus support (#6491) * Add Command R Plus GGUF * Add Command R Plus GGUF * Loading works up to LayerNorm2D * Export new tensors in 1D so they are not quantized. * Fix embedding layer based on Noeda's example * Whitespace * Add line * Fix unexpected tokens on MPS. Re-add F16 fix. ((Noeda) * dranger003: Fix block index overflow in CUDA dequantizing. * Reverted blocked multiplication code as it still has issues and could affect other Llama arches * export norms as f32 * fix overflow issues during quant and other cleanup * Type convention Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * dranger003: Fix more int overflow during quant. --------- Co-authored-by: S <seast@Ss-Mac-Studio.local> Co-authored-by: S <s@example.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-09 08:16:13 +00:00
if (n_layer >= 64){
layer.attn_q_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {hparams.n_embd_head_k, hparams.n_head});
layer.attn_k_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {hparams.n_embd_head_k, hparams.n_head_kv});
}
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd});
layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa});
layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
Implement the OLMo architecture (#6741) * implement olmo architecture * remove unused variable * remove unused moe branch * remove check for weight * remove superfluous moe, bias and rope tensors * clarified comment * fix clamp_kqv setting * remove obsolete parameter name filter
2024-04-19 09:35:54 +00:00
layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff});
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd});
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
}
} break;
case LLM_ARCH_OLMO: // adapted from LLM_ARCH_LLAMA with norm params removed
{
model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
// output
{
model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, false);
// if output is NULL, init from the input tok embed
if (model.output == NULL) {
model.output = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab});
ml.n_created--; // artificial tensor
ml.size_data += ggml_nbytes(model.output);
}
}
for (int i = 0; i < n_layer; ++i) {
ggml_context * ctx_split = ctx_for_layer_split(i);
auto & layer = model.layers[i];
layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd});
layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa});
layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa});
layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd});
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff});
layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd});
layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff});
}
} break;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
default:
throw std::runtime_error("unknown architecture");
build : enable more non-default compiler warnings (#3200)
2023-09-28 21:41:44 +00:00
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
ml.done_getting_tensors();
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
[Model] Add support for xverse (#6301) * Support xverse model convert to gguf format. * 1. Convert xverse models to gguf; 2. Add LLM_ARCH_XVERSE inference in llama.cpp; 3. Add xverse item in Supported models in README.md; * * gguf-py: remove redundant logs * llama: remove the init_mapping_prefetch custom parameter * llama.cpp: Include the changes from #6122 to exclude the unused outputs of the last layers. * - Fix format issues - Remove duplicate set kqv_out to llm_build_kv * Update llama.cpp --------- Co-authored-by: willhe <willhe@xverse.cn> Co-authored-by: willhe <hexin@xverse.cn>
2024-03-29 13:37:03 +00:00
ml.init_mappings(true, use_mlock ? &model.mlock_mmaps : nullptr);
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
model.mappings.reserve(ml.mappings.size());
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// create the backend buffers
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
std::vector<std::pair<ggml_context *, llama_buf_map>> ctx_bufs;
ctx_bufs.reserve(ctx_map.size());
// Ensure we have enough capacity for the maximum backend buffer we will potentially create
size_t n_max_backend_buffer = ctx_map.size() * ml.files.size();
model.bufs.reserve(n_max_backend_buffer);
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (auto & it : ctx_map) {
ggml_backend_buffer_type_t buft = it.first;
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
ggml_context * ctx = it.second;
llama_buf_map bufs;
bufs.reserve(n_max_backend_buffer);
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// only the mmap region containing the tensors in the model is mapped to the backend buffer
// this is important for metal with apple silicon: if the entire model could be mapped to a metal buffer, then we could just use metal for all layers
// this allows using partial offloading when the model size exceeds the metal buffer size, but not the RAM size
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
if (ml.use_mmap && use_mmap_buffer && buft == llama_default_buffer_type_cpu(true)) {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
for (uint32_t idx = 0; idx < ml.files.size(); idx++) {
void * addr = nullptr;
size_t first, last;
ml.get_mapping_range(&first, &last, &addr, idx, ctx);
if (first >= last) {
continue;
}
ggml_backend_buffer_t buf = ggml_backend_cpu_buffer_from_ptr((char *) addr + first, last - first);
if (buf == nullptr) {
throw std::runtime_error("unable to allocate backend CPU buffer");
}
model.bufs.push_back(buf);
bufs.emplace(idx, buf);
cuda : rename build flag to LLAMA_CUDA (#6299)
2024-03-26 00:16:01 +00:00
#ifdef GGML_USE_CUDA
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
if (n_layer >= n_gpu_layers) {
ggml_backend_cuda_register_host_buffer(
backend : offload large batches to GPU (#6083) * backend : offload large batches to GPU * fix hip * code cleanup * fix CUDA split buffers * Update ggml-backend-impl.h Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix memset without set_device * imatrix : remove sched affix from weight names * sched : add a new split if the current one has too many inputs reduce max inputs per split more cleanup * update backends ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-03-18 10:03:04 +00:00
ggml_backend_buffer_get_base(buf),
ggml_backend_buffer_get_size(buf));
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
}
backend : offload large batches to GPU (#6083) * backend : offload large batches to GPU * fix hip * code cleanup * fix CUDA split buffers * Update ggml-backend-impl.h Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix memset without set_device * imatrix : remove sched affix from weight names * sched : add a new split if the current one has too many inputs reduce max inputs per split more cleanup * update backends ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-03-18 10:03:04 +00:00
#endif
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
}
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
#ifdef GGML_USE_METAL
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
else if (ml.use_mmap && use_mmap_buffer && buft == ggml_backend_metal_buffer_type()) {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
for (uint32_t idx = 0; idx < ml.files.size(); idx++) {
const size_t max_size = ggml_get_max_tensor_size(ctx);
void * addr = nullptr;
size_t first, last;
ml.get_mapping_range(&first, &last, &addr, idx, ctx);
if (first >= last) {
continue;
}
ggml_backend_buffer_t buf = ggml_backend_metal_buffer_from_ptr((char *) addr + first, last - first, max_size);
if (buf == nullptr) {
throw std::runtime_error("unable to allocate backend metal buffer");
}
model.bufs.push_back(buf);
bufs.emplace(idx, buf);
}
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
#endif
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
else {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
ggml_backend_buffer_t buf = ggml_backend_alloc_ctx_tensors_from_buft(ctx, buft);
if (buf == nullptr) {
throw std::runtime_error("unable to allocate backend buffer");
}
model.bufs.push_back(buf);
if (use_mlock && ggml_backend_buffer_is_host(buf)) {
llama : fix mlock with no-mmap with Metal (#5025)
2024-01-18 20:12:15 +00:00
model.mlock_bufs.emplace_back(new llama_mlock);
auto & mlock_buf = model.mlock_bufs.back();
mlock_buf->init (ggml_backend_buffer_get_base(buf));
mlock_buf->grow_to(ggml_backend_buffer_get_size(buf));
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
for (uint32_t idx = 0; idx < ml.files.size(); idx++) {
bufs.emplace(idx, buf);
}
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
if (bufs.empty()) {
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
throw std::runtime_error("failed to allocate buffer");
}
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
for (auto & buf : bufs) {
// indicate that this buffer contains weights
// this is used by ggml_backend_sched to improve op scheduling -> ops that use a weight are preferably scheduled to the backend that contains the weight
ggml_backend_buffer_set_usage(buf.second, GGML_BACKEND_BUFFER_USAGE_WEIGHTS);
}
ctx_bufs.emplace_back(ctx, bufs);
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
llama : do not print "offloading layers" message in CPU-only builds (#5416)
2024-02-08 20:33:03 +00:00
if (llama_supports_gpu_offload()) {
llama : fix compile warnings
2023-06-06 19:41:53 +00:00
const int n_gpu = std::min(n_gpu_layers, int(hparams.n_layer));
add log_callback to llama_context_params for custom logging. (#2234) * add log_callback to llama_context_params for custom logging. * Fix macro expansion on gcc * Add struct llama_state for global variables and move log_callback there * Turn log level into enum and some minor changes. * Remove model_for_logging parameter (not needed anymore) * Convert remaining fprintf(stderr, ...) calls to use new macros. * Fix enum and initialize g_state * Fix log calls after merge * Fix missing static * Add back all the new lines in the logging strings * Add comment for llama_log_callback and replace remaining printf calls --------- Co-authored-by: grahameth <-> Co-authored-by: Helmut <helmut.buhler@inf.h-brs.de>
2023-08-09 20:46:40 +00:00
LLAMA_LOG_INFO("%s: offloading %d repeating layers to GPU\n", __func__, n_gpu);
cuda : loading models directly into VRAM, norm calculation on GPU, broadcasting for ggml_mul (#1483) * Broadcasting for ggml_mul * CUDA kernel for ggml_mul, norms in VRAM * GPU weights not in RAM, direct loading with cuFile * fixup! GPU weights not in RAM, direct loading with cuFile * fixup! GPU weights not in RAM, direct loading with cuFile * define default model path once, sync path with readme (#1366) * ~7% faster Q5_1 AVX2 code (#1477) * convert.py: Support models which are stored in a single pytorch_model.bin (#1469) * Support models in a single pytorch_model.bin * Remove spurious line with typo * benchmark-matmul: Print the average of the test results (#1490) * Remove unused n_parts parameter (#1509) * Fixes #1511 lambda issue for w64devkit (mingw) (#1513) * Fix for w64devkit and mingw * make kv_f16 the default for api users (#1517) * minor : fix compile warnings * readme : adds WizardLM to the list of supported models (#1485) * main : make reverse prompt option act as a stop token in non-interactive mode (#1032) * Make reverse prompt option act as a stop token in non-interactive scenarios * Making requested review changes * Update gpt_params_parse and fix a merge error * Revert "Update gpt_params_parse and fix a merge error" This reverts commit 2bb2ff1748513591ad45b175a75ed1d8089d84c8. * Update gpt_params_parse and fix a merge error take 2 * examples : add persistent chat (#1495) * examples : add persistent chat * examples : fix whitespace --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * tests : add missing header * ggml : use F16 instead of F32 in Q4_0, Q4_1, Q8_0 (#1508) * ggml : use F16 instead of F32 in Q4_0, Q4_1 and Q8_0 * llama : bump LLAMA_FILE_VERSION to 3 * cuda : update Q4 and Q8 dequantize kernels * ggml : fix AVX dot products * readme : update performance table + hot topics * ggml : fix scalar implementation of Q4_1 dot * llama : fix compile warnings in llama_set_state_data() * llama : fix name shadowing and C4146 (#1526) * Fix name shadowing and C4146 * Fix if macros not using defined when required * Update llama-util.h Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Update llama-util.h Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Code style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix for mingw (#1462) * llama : add llama_init_backend() API (close #1527) * feature : add blis and other BLAS implementation support (#1502) * feature: add blis support * feature: allow all BLA_VENDOR to be assigned in cmake arguments. align with whisper.cpp pr 927 * fix: version detection for BLA_SIZEOF_INTEGER, recover min version of cmake * Fix typo in INTEGER Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Revert "feature : add blis and other BLAS implementation support (#1502)" This reverts commit 07e9ace0f9da424d82e75df969642522880feb92. * GPU weights not in RAM, direct loading with cuFile * llama : code style fixes + progress print fix * ggml : ggml_mul better broadcast support * cmake : workarounds for cufile when CMake version < 3.25 * gg rebase fixup * Loop in llama.cpp, fixed progress callback * Attempt clang-tidy fix * llama : fix vram size computation * Add forgotten fclose() --------- Co-authored-by: András Salamon <ott2@users.noreply.github.com> Co-authored-by: Ilya Kurdyukov <59548320+ilyakurdyukov@users.noreply.github.com> Co-authored-by: Tom Jobbins <784313+TheBloke@users.noreply.github.com> Co-authored-by: rankaiyx <rankaiyx@rankaiyx.com> Co-authored-by: Stephan Walter <stephan@walter.name> Co-authored-by: DannyDaemonic <DannyDaemonic@gmail.com> Co-authored-by: Erik Scholz <Green-Sky@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: David Kennedy <dakennedyd@gmail.com> Co-authored-by: Jason McCartney <jmac@theroot.org> Co-authored-by: Evan Jones <evan.q.jones@gmail.com> Co-authored-by: Maxime <672982+maximegmd@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Zenix <zenixls2@gmail.com>
2023-05-20 12:19:28 +00:00
if (n_gpu_layers > (int) hparams.n_layer) {
add log_callback to llama_context_params for custom logging. (#2234) * add log_callback to llama_context_params for custom logging. * Fix macro expansion on gcc * Add struct llama_state for global variables and move log_callback there * Turn log level into enum and some minor changes. * Remove model_for_logging parameter (not needed anymore) * Convert remaining fprintf(stderr, ...) calls to use new macros. * Fix enum and initialize g_state * Fix log calls after merge * Fix missing static * Add back all the new lines in the logging strings * Add comment for llama_log_callback and replace remaining printf calls --------- Co-authored-by: grahameth <-> Co-authored-by: Helmut <helmut.buhler@inf.h-brs.de>
2023-08-09 20:46:40 +00:00
LLAMA_LOG_INFO("%s: offloading non-repeating layers to GPU\n", __func__);
CUDA full GPU acceleration, KV cache in VRAM (#1827) * Fixed CUDA RoPE * ggml_cuda_mul_mat_vec_p021 * ggml_cuda_scale * ggml_cuda_diag_mask_inf * ggml_is_permuted * ggml_cuda_cpy * flatten rows for ggml_cuda_op * Added a --low-vram option * Fixed Windows performance * Fixed LLAMA_CUDA_DMMV_Y > 1 for WizardLM
2023-06-14 17:47:19 +00:00
}
Fixed OpenCL offloading prints (#2082)
2023-07-05 06:58:05 +00:00
const int max_backend_supported_layers = hparams.n_layer + 1;
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
const int max_offloadable_layers = hparams.n_layer + 1;
Fixed OpenCL offloading prints (#2082)
2023-07-05 06:58:05 +00:00
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
LLAMA_LOG_INFO("%s: offloaded %d/%d layers to GPU\n", __func__, std::min(n_gpu_layers, max_offloadable_layers), max_backend_supported_layers);
llama : do not print "offloading layers" message in CPU-only builds (#5416)
2024-02-08 20:33:03 +00:00
}
ggml : GPU-accelerated token generation (#1412) * CUDA kernel for q4_0 dequant. + mat. vec. mult. * Added q4_1 via template * Added missing __syncthreads(); * --gpu_layers -> --gpu-layers * Shorter dequantize_mul_mat_vec line * q5_0 dequantize_mul_mat kernel * More readable dequantize_mul_mat_vec logic * dequantize_mul_mat_vec kernels for q5_1, q8_0, f16 * llama : offload "output" tensor to GPU too + coding style fixes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-05-13 13:38:36 +00:00
llama : do not print "offloading layers" message in CPU-only builds (#5416)
2024-02-08 20:33:03 +00:00
// print memory requirements
for (ggml_backend_buffer_t buf : model.bufs) {
LLAMA_LOG_INFO("%s: %10s buffer size = %8.2f MiB\n", __func__, ggml_backend_buffer_name(buf), ggml_backend_buffer_get_size(buf) / 1024.0 / 1024.0);
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
}
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
// populate tensors_by_name
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (ggml_context * ctx : model.ctxs) {
for (auto * cur = ggml_get_first_tensor(ctx); cur != NULL; cur = ggml_get_next_tensor(ctx, cur)) {
model.tensors_by_name.emplace_back(ggml_get_name(cur), cur);
}
cuda : loading models directly into VRAM, norm calculation on GPU, broadcasting for ggml_mul (#1483) * Broadcasting for ggml_mul * CUDA kernel for ggml_mul, norms in VRAM * GPU weights not in RAM, direct loading with cuFile * fixup! GPU weights not in RAM, direct loading with cuFile * fixup! GPU weights not in RAM, direct loading with cuFile * define default model path once, sync path with readme (#1366) * ~7% faster Q5_1 AVX2 code (#1477) * convert.py: Support models which are stored in a single pytorch_model.bin (#1469) * Support models in a single pytorch_model.bin * Remove spurious line with typo * benchmark-matmul: Print the average of the test results (#1490) * Remove unused n_parts parameter (#1509) * Fixes #1511 lambda issue for w64devkit (mingw) (#1513) * Fix for w64devkit and mingw * make kv_f16 the default for api users (#1517) * minor : fix compile warnings * readme : adds WizardLM to the list of supported models (#1485) * main : make reverse prompt option act as a stop token in non-interactive mode (#1032) * Make reverse prompt option act as a stop token in non-interactive scenarios * Making requested review changes * Update gpt_params_parse and fix a merge error * Revert "Update gpt_params_parse and fix a merge error" This reverts commit 2bb2ff1748513591ad45b175a75ed1d8089d84c8. * Update gpt_params_parse and fix a merge error take 2 * examples : add persistent chat (#1495) * examples : add persistent chat * examples : fix whitespace --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * tests : add missing header * ggml : use F16 instead of F32 in Q4_0, Q4_1, Q8_0 (#1508) * ggml : use F16 instead of F32 in Q4_0, Q4_1 and Q8_0 * llama : bump LLAMA_FILE_VERSION to 3 * cuda : update Q4 and Q8 dequantize kernels * ggml : fix AVX dot products * readme : update performance table + hot topics * ggml : fix scalar implementation of Q4_1 dot * llama : fix compile warnings in llama_set_state_data() * llama : fix name shadowing and C4146 (#1526) * Fix name shadowing and C4146 * Fix if macros not using defined when required * Update llama-util.h Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Update llama-util.h Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Code style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix for mingw (#1462) * llama : add llama_init_backend() API (close #1527) * feature : add blis and other BLAS implementation support (#1502) * feature: add blis support * feature: allow all BLA_VENDOR to be assigned in cmake arguments. align with whisper.cpp pr 927 * fix: version detection for BLA_SIZEOF_INTEGER, recover min version of cmake * Fix typo in INTEGER Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Revert "feature : add blis and other BLAS implementation support (#1502)" This reverts commit 07e9ace0f9da424d82e75df969642522880feb92. * GPU weights not in RAM, direct loading with cuFile * llama : code style fixes + progress print fix * ggml : ggml_mul better broadcast support * cmake : workarounds for cufile when CMake version < 3.25 * gg rebase fixup * Loop in llama.cpp, fixed progress callback * Attempt clang-tidy fix * llama : fix vram size computation * Add forgotten fclose() --------- Co-authored-by: András Salamon <ott2@users.noreply.github.com> Co-authored-by: Ilya Kurdyukov <59548320+ilyakurdyukov@users.noreply.github.com> Co-authored-by: Tom Jobbins <784313+TheBloke@users.noreply.github.com> Co-authored-by: rankaiyx <rankaiyx@rankaiyx.com> Co-authored-by: Stephan Walter <stephan@walter.name> Co-authored-by: DannyDaemonic <DannyDaemonic@gmail.com> Co-authored-by: Erik Scholz <Green-Sky@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: David Kennedy <dakennedyd@gmail.com> Co-authored-by: Jason McCartney <jmac@theroot.org> Co-authored-by: Evan Jones <evan.q.jones@gmail.com> Co-authored-by: Maxime <672982+maximegmd@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Zenix <zenixls2@gmail.com>
2023-05-20 12:19:28 +00:00
}
ggml : GPU-accelerated token generation (#1412) * CUDA kernel for q4_0 dequant. + mat. vec. mult. * Added q4_1 via template * Added missing __syncthreads(); * --gpu_layers -> --gpu-layers * Shorter dequantize_mul_mat_vec line * q5_0 dequantize_mul_mat kernel * More readable dequantize_mul_mat_vec logic * dequantize_mul_mat_vec kernels for q5_1, q8_0, f16 * llama : offload "output" tensor to GPU too + coding style fixes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-05-13 13:38:36 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// load tensor data
for (auto & it : ctx_bufs) {
ggml_context * ctx = it.first;
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
auto & bufs = it.second;
if (!ml.load_all_data(ctx, bufs, use_mlock ? &model.mlock_mmaps : NULL, progress_callback, progress_callback_user_data)) {
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
return false;
}
llama : add ability to cancel model loading (#4462) * llama : Add ability to cancel model load Updated llama_progress_callback so that if it returns false, the model loading is aborted. * llama : Add test for model load cancellation * Fix bool return in llama_model_load, remove std::ignore use * Update llama.cpp Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Fail test if model file is missing * Revert "Fail test if model file is missing" This reverts commit 32ebd525bf7e5a87ee8a3dbaab3d92ce79fbf23d. * Add test-model-load-cancel to Makefile * Revert "Revert "Fail test if model file is missing"" This reverts commit 2796953257ee5383fa7c8fe8fa8fc888c048fb0b. * Simplify .gitignore for tests, clang-tidy fixes * Label all ctest tests * ci : ctest uses -L main * Attempt at writing ctest_with_model * ci : get ci/run.sh working with test-model-load-cancel * ci : restrict .github/workflows/build.yml ctest to -L main * update requirements.txt * Disable test-model-load-cancel in make * Remove venv before creation * Restructure requirements.txt Top-level now imports the specific additional requirements for each python file. Using `pip install -r requirements.txt` will fail if versions become mismatched in the per-file requirements. * Make per-python-script requirements work alone This doesn't break the main requirements.txt. * Add comment * Add convert-persimmon-to-gguf.py to new requirements.txt scheme * Add check-requirements.sh script and GitHub workflow * Remove shellcheck installation step from workflow * Add nocleanup special arg * Fix merge see: https://github.com/ggerganov/llama.cpp/pull/4462#discussion_r1434593573 * reset to upstream/master * Redo changes for cancelling model load --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2023-12-22 06:19:36 +00:00
}
cuda : loading models directly into VRAM, norm calculation on GPU, broadcasting for ggml_mul (#1483) * Broadcasting for ggml_mul * CUDA kernel for ggml_mul, norms in VRAM * GPU weights not in RAM, direct loading with cuFile * fixup! GPU weights not in RAM, direct loading with cuFile * fixup! GPU weights not in RAM, direct loading with cuFile * define default model path once, sync path with readme (#1366) * ~7% faster Q5_1 AVX2 code (#1477) * convert.py: Support models which are stored in a single pytorch_model.bin (#1469) * Support models in a single pytorch_model.bin * Remove spurious line with typo * benchmark-matmul: Print the average of the test results (#1490) * Remove unused n_parts parameter (#1509) * Fixes #1511 lambda issue for w64devkit (mingw) (#1513) * Fix for w64devkit and mingw * make kv_f16 the default for api users (#1517) * minor : fix compile warnings * readme : adds WizardLM to the list of supported models (#1485) * main : make reverse prompt option act as a stop token in non-interactive mode (#1032) * Make reverse prompt option act as a stop token in non-interactive scenarios * Making requested review changes * Update gpt_params_parse and fix a merge error * Revert "Update gpt_params_parse and fix a merge error" This reverts commit 2bb2ff1748513591ad45b175a75ed1d8089d84c8. * Update gpt_params_parse and fix a merge error take 2 * examples : add persistent chat (#1495) * examples : add persistent chat * examples : fix whitespace --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * tests : add missing header * ggml : use F16 instead of F32 in Q4_0, Q4_1, Q8_0 (#1508) * ggml : use F16 instead of F32 in Q4_0, Q4_1 and Q8_0 * llama : bump LLAMA_FILE_VERSION to 3 * cuda : update Q4 and Q8 dequantize kernels * ggml : fix AVX dot products * readme : update performance table + hot topics * ggml : fix scalar implementation of Q4_1 dot * llama : fix compile warnings in llama_set_state_data() * llama : fix name shadowing and C4146 (#1526) * Fix name shadowing and C4146 * Fix if macros not using defined when required * Update llama-util.h Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Update llama-util.h Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> * Code style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix for mingw (#1462) * llama : add llama_init_backend() API (close #1527) * feature : add blis and other BLAS implementation support (#1502) * feature: add blis support * feature: allow all BLA_VENDOR to be assigned in cmake arguments. align with whisper.cpp pr 927 * fix: version detection for BLA_SIZEOF_INTEGER, recover min version of cmake * Fix typo in INTEGER Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Revert "feature : add blis and other BLAS implementation support (#1502)" This reverts commit 07e9ace0f9da424d82e75df969642522880feb92. * GPU weights not in RAM, direct loading with cuFile * llama : code style fixes + progress print fix * ggml : ggml_mul better broadcast support * cmake : workarounds for cufile when CMake version < 3.25 * gg rebase fixup * Loop in llama.cpp, fixed progress callback * Attempt clang-tidy fix * llama : fix vram size computation * Add forgotten fclose() --------- Co-authored-by: András Salamon <ott2@users.noreply.github.com> Co-authored-by: Ilya Kurdyukov <59548320+ilyakurdyukov@users.noreply.github.com> Co-authored-by: Tom Jobbins <784313+TheBloke@users.noreply.github.com> Co-authored-by: rankaiyx <rankaiyx@rankaiyx.com> Co-authored-by: Stephan Walter <stephan@walter.name> Co-authored-by: DannyDaemonic <DannyDaemonic@gmail.com> Co-authored-by: Erik Scholz <Green-Sky@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: David Kennedy <dakennedyd@gmail.com> Co-authored-by: Jason McCartney <jmac@theroot.org> Co-authored-by: Evan Jones <evan.q.jones@gmail.com> Co-authored-by: Maxime <672982+maximegmd@users.noreply.github.com> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: Zenix <zenixls2@gmail.com>
2023-05-20 12:19:28 +00:00
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
if (use_mmap_buffer) {
for (auto & mapping : ml.mappings) {
model.mappings.emplace_back(std::move(mapping));
}
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
Make loading weights 10-100x faster This is a breaking change that's going to give you three benefits: 1. Your inference commands should load 100x faster 2. You may be able to safely load models 2x larger 3. You can run many concurrent inference processes This was accomplished by changing the file format so we can mmap() weights directly into memory without having to read() or copy them thereby ensuring the kernel can make its file cache pages directly accessible to our inference processes; and secondly, that the file cache pages are much less likely to get evicted (which would force loads to hit disk) because they're no longer competing with memory pages that were needlessly created by gigabytes of standard i/o. The new file format supports single-file models like LLaMA 7b, and it also supports multi-file models like LLaMA 13B. Our Python tool now merges the foo.1, foo.2, etc. files back into a single file so that the C++ code which maps it doesn't need to reshape data every time. That's made llama.cpp so much simpler. Much of its load code has now been deleted. Furthermore, this change ensures that tensors are aligned properly on a 32-byte boundary. That opens the door to seeing if we can get additional performance gains on some microprocessors, by using ops that require memory alignment. Lastly note that both POSIX and the Windows platform are supported Fixes #91
2023-03-29 20:51:37 +00:00
// loading time will be recalculate after the first eval, so
// we take page faults deferred by mmap() into consideration
llama : make model stateless and context stateful (llama_state) (#1797) * llama : make model stateless and context stateful * llama : minor cleanup * llama : update internal API declaration * Apply suggestions from code review fix style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Missing model memory release * Fix style * Add deprecated warning for public API function llama_init_from_file * Update public API use cases: move away from deprecated llama_init_from_file * Deprecate public API function llama_apply_lora_from_file --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-24 08:47:58 +00:00
model.t_load_us = ggml_time_us() - model.t_start_us;
llama : add ability to cancel model loading (#4462) * llama : Add ability to cancel model load Updated llama_progress_callback so that if it returns false, the model loading is aborted. * llama : Add test for model load cancellation * Fix bool return in llama_model_load, remove std::ignore use * Update llama.cpp Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Fail test if model file is missing * Revert "Fail test if model file is missing" This reverts commit 32ebd525bf7e5a87ee8a3dbaab3d92ce79fbf23d. * Add test-model-load-cancel to Makefile * Revert "Revert "Fail test if model file is missing"" This reverts commit 2796953257ee5383fa7c8fe8fa8fc888c048fb0b. * Simplify .gitignore for tests, clang-tidy fixes * Label all ctest tests * ci : ctest uses -L main * Attempt at writing ctest_with_model * ci : get ci/run.sh working with test-model-load-cancel * ci : restrict .github/workflows/build.yml ctest to -L main * update requirements.txt * Disable test-model-load-cancel in make * Remove venv before creation * Restructure requirements.txt Top-level now imports the specific additional requirements for each python file. Using `pip install -r requirements.txt` will fail if versions become mismatched in the per-file requirements. * Make per-python-script requirements work alone This doesn't break the main requirements.txt. * Add comment * Add convert-persimmon-to-gguf.py to new requirements.txt scheme * Add check-requirements.sh script and GitHub workflow * Remove shellcheck installation step from workflow * Add nocleanup special arg * Fix merge see: https://github.com/ggerganov/llama.cpp/pull/4462#discussion_r1434593573 * reset to upstream/master * Redo changes for cancelling model load --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2023-12-22 06:19:36 +00:00
return true;
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : add ability to cancel model loading (#4462) * llama : Add ability to cancel model load Updated llama_progress_callback so that if it returns false, the model loading is aborted. * llama : Add test for model load cancellation * Fix bool return in llama_model_load, remove std::ignore use * Update llama.cpp Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Fail test if model file is missing * Revert "Fail test if model file is missing" This reverts commit 32ebd525bf7e5a87ee8a3dbaab3d92ce79fbf23d. * Add test-model-load-cancel to Makefile * Revert "Revert "Fail test if model file is missing"" This reverts commit 2796953257ee5383fa7c8fe8fa8fc888c048fb0b. * Simplify .gitignore for tests, clang-tidy fixes * Label all ctest tests * ci : ctest uses -L main * Attempt at writing ctest_with_model * ci : get ci/run.sh working with test-model-load-cancel * ci : restrict .github/workflows/build.yml ctest to -L main * update requirements.txt * Disable test-model-load-cancel in make * Remove venv before creation * Restructure requirements.txt Top-level now imports the specific additional requirements for each python file. Using `pip install -r requirements.txt` will fail if versions become mismatched in the per-file requirements. * Make per-python-script requirements work alone This doesn't break the main requirements.txt. * Add comment * Add convert-persimmon-to-gguf.py to new requirements.txt scheme * Add check-requirements.sh script and GitHub workflow * Remove shellcheck installation step from workflow * Add nocleanup special arg * Fix merge see: https://github.com/ggerganov/llama.cpp/pull/4462#discussion_r1434593573 * reset to upstream/master * Redo changes for cancelling model load --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2023-12-22 06:19:36 +00:00
// Returns 0 on success, -1 on error, and -2 on cancellation via llama_progress_callback
Nomic Vulkan backend (#4456) Signed-off-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: niansa <anton-sa@web.de> Co-authored-by: Adam Treat <treat.adam@gmail.com> Co-authored-by: Aaron Miller <apage43@ninjawhale.com> Co-authored-by: ToKiNoBug <tokinobug@163.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-01-29 20:50:50 +00:00
static int llama_model_load(const std::string & fname, llama_model & model, llama_model_params & params) {
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
try {
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
llama_model_loader ml(fname, params.use_mmap, params.check_tensors, params.kv_overrides);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
model.hparams.vocab_only = params.vocab_only;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : make load error reporting more granular (#5477) Makes it easier to pinpoint where e.g. `unordered_map::at: key not found` comes from.
2024-02-13 13:24:50 +00:00
try {
llm_load_arch(ml, model);
} catch(const std::exception & e) {
throw std::runtime_error("error loading model architecture: " + std::string(e.what()));
}
try {
llm_load_hparams(ml, model);
} catch(const std::exception & e) {
throw std::runtime_error("error loading model hyperparameters: " + std::string(e.what()));
}
try {
llm_load_vocab(ml, model);
} catch(const std::exception & e) {
throw std::runtime_error("error loading model vocabulary: " + std::string(e.what()));
}
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
llm_load_print_meta(ml, model);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
if (model.vocab.type != LLAMA_VOCAB_TYPE_NONE &&
model.hparams.n_vocab != model.vocab.id_to_token.size()) {
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
throw std::runtime_error("vocab size mismatch");
}
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
if (params.vocab_only) {
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
LLAMA_LOG_INFO("%s: vocab only - skipping tensors\n", __func__);
llama : add ability to cancel model loading (#4462) * llama : Add ability to cancel model load Updated llama_progress_callback so that if it returns false, the model loading is aborted. * llama : Add test for model load cancellation * Fix bool return in llama_model_load, remove std::ignore use * Update llama.cpp Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Fail test if model file is missing * Revert "Fail test if model file is missing" This reverts commit 32ebd525bf7e5a87ee8a3dbaab3d92ce79fbf23d. * Add test-model-load-cancel to Makefile * Revert "Revert "Fail test if model file is missing"" This reverts commit 2796953257ee5383fa7c8fe8fa8fc888c048fb0b. * Simplify .gitignore for tests, clang-tidy fixes * Label all ctest tests * ci : ctest uses -L main * Attempt at writing ctest_with_model * ci : get ci/run.sh working with test-model-load-cancel * ci : restrict .github/workflows/build.yml ctest to -L main * update requirements.txt * Disable test-model-load-cancel in make * Remove venv before creation * Restructure requirements.txt Top-level now imports the specific additional requirements for each python file. Using `pip install -r requirements.txt` will fail if versions become mismatched in the per-file requirements. * Make per-python-script requirements work alone This doesn't break the main requirements.txt. * Add comment * Add convert-persimmon-to-gguf.py to new requirements.txt scheme * Add check-requirements.sh script and GitHub workflow * Remove shellcheck installation step from workflow * Add nocleanup special arg * Fix merge see: https://github.com/ggerganov/llama.cpp/pull/4462#discussion_r1434593573 * reset to upstream/master * Redo changes for cancelling model load --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2023-12-22 06:19:36 +00:00
return 0;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
Nomic Vulkan backend (#4456) Signed-off-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: niansa <anton-sa@web.de> Co-authored-by: Adam Treat <treat.adam@gmail.com> Co-authored-by: Aaron Miller <apage43@ninjawhale.com> Co-authored-by: ToKiNoBug <tokinobug@163.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-01-29 20:50:50 +00:00
#ifdef GGML_USE_KOMPUTE
kompute : fix fallback to CPU (#5201)
2024-01-29 22:11:27 +00:00
if (params.n_gpu_layers > 0 && (
Nomic Vulkan backend (#4456) Signed-off-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: niansa <anton-sa@web.de> Co-authored-by: Adam Treat <treat.adam@gmail.com> Co-authored-by: Aaron Miller <apage43@ninjawhale.com> Co-authored-by: ToKiNoBug <tokinobug@163.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-01-29 20:50:50 +00:00
!(model.arch == LLM_ARCH_LLAMA || model.arch == LLM_ARCH_FALCON)
|| !(
model.ftype == LLAMA_FTYPE_ALL_F32 ||
model.ftype == LLAMA_FTYPE_MOSTLY_F16 ||
model.ftype == LLAMA_FTYPE_MOSTLY_Q4_0 ||
model.ftype == LLAMA_FTYPE_MOSTLY_Q4_1
)
)) {
// TODO(cebtenzzre): propagate this error outside of llama_load_model_from_file
kompute : fix fallback to CPU (#5201)
2024-01-29 22:11:27 +00:00
LLAMA_LOG_WARN("%s: disabling Kompute due to unsupported model arch or quantization\n", __func__);
Nomic Vulkan backend (#4456) Signed-off-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: niansa <anton-sa@web.de> Co-authored-by: Adam Treat <treat.adam@gmail.com> Co-authored-by: Aaron Miller <apage43@ninjawhale.com> Co-authored-by: ToKiNoBug <tokinobug@163.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-01-29 20:50:50 +00:00
params.n_gpu_layers = 0;
}
#endif
fix set main gpu error (#6073)
2024-03-15 10:53:53 +00:00
#ifdef GGML_USE_SYCL
if (params.split_mode == LLAMA_SPLIT_MODE_NONE) {
ggml_backend_sycl_set_single_device_mode(params.main_gpu);
//SYCL use device index (0, 1, 2) directly, uer input device id, then convert to device index.
params.main_gpu = ggml_backend_sycl_get_device_index(params.main_gpu);
} else {
ggml_backend_sycl_set_mul_device_mode();
}
#endif
llama : add ability to cancel model loading (#4462) * llama : Add ability to cancel model load Updated llama_progress_callback so that if it returns false, the model loading is aborted. * llama : Add test for model load cancellation * Fix bool return in llama_model_load, remove std::ignore use * Update llama.cpp Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Fail test if model file is missing * Revert "Fail test if model file is missing" This reverts commit 32ebd525bf7e5a87ee8a3dbaab3d92ce79fbf23d. * Add test-model-load-cancel to Makefile * Revert "Revert "Fail test if model file is missing"" This reverts commit 2796953257ee5383fa7c8fe8fa8fc888c048fb0b. * Simplify .gitignore for tests, clang-tidy fixes * Label all ctest tests * ci : ctest uses -L main * Attempt at writing ctest_with_model * ci : get ci/run.sh working with test-model-load-cancel * ci : restrict .github/workflows/build.yml ctest to -L main * update requirements.txt * Disable test-model-load-cancel in make * Remove venv before creation * Restructure requirements.txt Top-level now imports the specific additional requirements for each python file. Using `pip install -r requirements.txt` will fail if versions become mismatched in the per-file requirements. * Make per-python-script requirements work alone This doesn't break the main requirements.txt. * Add comment * Add convert-persimmon-to-gguf.py to new requirements.txt scheme * Add check-requirements.sh script and GitHub workflow * Remove shellcheck installation step from workflow * Add nocleanup special arg * Fix merge see: https://github.com/ggerganov/llama.cpp/pull/4462#discussion_r1434593573 * reset to upstream/master * Redo changes for cancelling model load --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2023-12-22 06:19:36 +00:00
if (!llm_load_tensors(
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ml, model, params.n_gpu_layers, params.split_mode, params.main_gpu, params.tensor_split, params.use_mlock,
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
params.progress_callback, params.progress_callback_user_data
llama : add ability to cancel model loading (#4462) * llama : Add ability to cancel model load Updated llama_progress_callback so that if it returns false, the model loading is aborted. * llama : Add test for model load cancellation * Fix bool return in llama_model_load, remove std::ignore use * Update llama.cpp Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Fail test if model file is missing * Revert "Fail test if model file is missing" This reverts commit 32ebd525bf7e5a87ee8a3dbaab3d92ce79fbf23d. * Add test-model-load-cancel to Makefile * Revert "Revert "Fail test if model file is missing"" This reverts commit 2796953257ee5383fa7c8fe8fa8fc888c048fb0b. * Simplify .gitignore for tests, clang-tidy fixes * Label all ctest tests * ci : ctest uses -L main * Attempt at writing ctest_with_model * ci : get ci/run.sh working with test-model-load-cancel * ci : restrict .github/workflows/build.yml ctest to -L main * update requirements.txt * Disable test-model-load-cancel in make * Remove venv before creation * Restructure requirements.txt Top-level now imports the specific additional requirements for each python file. Using `pip install -r requirements.txt` will fail if versions become mismatched in the per-file requirements. * Make per-python-script requirements work alone This doesn't break the main requirements.txt. * Add comment * Add convert-persimmon-to-gguf.py to new requirements.txt scheme * Add check-requirements.sh script and GitHub workflow * Remove shellcheck installation step from workflow * Add nocleanup special arg * Fix merge see: https://github.com/ggerganov/llama.cpp/pull/4462#discussion_r1434593573 * reset to upstream/master * Redo changes for cancelling model load --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2023-12-22 06:19:36 +00:00
)) {
return -2;
}
llama : consistently catch and throw only exceptions deriving from std::exception (#1599) Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-05 20:24:29 +00:00
} catch (const std::exception & err) {
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
LLAMA_LOG_ERROR("%s: error loading model: %s\n", __func__, err.what());
llama : add ability to cancel model loading (#4462) * llama : Add ability to cancel model load Updated llama_progress_callback so that if it returns false, the model loading is aborted. * llama : Add test for model load cancellation * Fix bool return in llama_model_load, remove std::ignore use * Update llama.cpp Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Fail test if model file is missing * Revert "Fail test if model file is missing" This reverts commit 32ebd525bf7e5a87ee8a3dbaab3d92ce79fbf23d. * Add test-model-load-cancel to Makefile * Revert "Revert "Fail test if model file is missing"" This reverts commit 2796953257ee5383fa7c8fe8fa8fc888c048fb0b. * Simplify .gitignore for tests, clang-tidy fixes * Label all ctest tests * ci : ctest uses -L main * Attempt at writing ctest_with_model * ci : get ci/run.sh working with test-model-load-cancel * ci : restrict .github/workflows/build.yml ctest to -L main * update requirements.txt * Disable test-model-load-cancel in make * Remove venv before creation * Restructure requirements.txt Top-level now imports the specific additional requirements for each python file. Using `pip install -r requirements.txt` will fail if versions become mismatched in the per-file requirements. * Make per-python-script requirements work alone This doesn't break the main requirements.txt. * Add comment * Add convert-persimmon-to-gguf.py to new requirements.txt scheme * Add check-requirements.sh script and GitHub workflow * Remove shellcheck installation step from workflow * Add nocleanup special arg * Fix merge see: https://github.com/ggerganov/llama.cpp/pull/4462#discussion_r1434593573 * reset to upstream/master * Redo changes for cancelling model load --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2023-12-22 06:19:36 +00:00
return -1;
Add support for file load progress reporting callbacks (#434) * File load progress reporting * Move llama_progress_handler into llama_context_params * Renames * Use seekg to find file size instead * More correct load progress * Call progress callback more frequently * Fix typo
2023-03-25 05:26:28 +00:00
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : add ability to cancel model loading (#4462) * llama : Add ability to cancel model load Updated llama_progress_callback so that if it returns false, the model loading is aborted. * llama : Add test for model load cancellation * Fix bool return in llama_model_load, remove std::ignore use * Update llama.cpp Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Fail test if model file is missing * Revert "Fail test if model file is missing" This reverts commit 32ebd525bf7e5a87ee8a3dbaab3d92ce79fbf23d. * Add test-model-load-cancel to Makefile * Revert "Revert "Fail test if model file is missing"" This reverts commit 2796953257ee5383fa7c8fe8fa8fc888c048fb0b. * Simplify .gitignore for tests, clang-tidy fixes * Label all ctest tests * ci : ctest uses -L main * Attempt at writing ctest_with_model * ci : get ci/run.sh working with test-model-load-cancel * ci : restrict .github/workflows/build.yml ctest to -L main * update requirements.txt * Disable test-model-load-cancel in make * Remove venv before creation * Restructure requirements.txt Top-level now imports the specific additional requirements for each python file. Using `pip install -r requirements.txt` will fail if versions become mismatched in the per-file requirements. * Make per-python-script requirements work alone This doesn't break the main requirements.txt. * Add comment * Add convert-persimmon-to-gguf.py to new requirements.txt scheme * Add check-requirements.sh script and GitHub workflow * Remove shellcheck installation step from workflow * Add nocleanup special arg * Fix merge see: https://github.com/ggerganov/llama.cpp/pull/4462#discussion_r1434593573 * reset to upstream/master * Redo changes for cancelling model load --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2023-12-22 06:19:36 +00:00
return 0;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
//
// llm_build
//
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
using llm_build_cb = std::function<void(struct ggml_tensor * cur, const char * name, int nl)>;
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
enum llm_ffn_op_type {
LLM_FFN_SILU,
LLM_FFN_GELU,
LLM_FFN_RELU,
LLM_FFN_RELU_SQR,
};
enum llm_ffn_gate_type {
LLM_FFN_SEQ,
LLM_FFN_PAR, // ffn_gate is parallel to ffn_up
};
enum llm_norm_type {
LLM_NORM,
LLM_NORM_RMS,
};
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
static struct ggml_tensor * llm_build_inp_embd(
struct ggml_context * ctx,
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct llama_context & lctx,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
const llama_hparams & hparams,
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
const llama_batch & batch,
struct ggml_tensor * tok_embd,
const llm_build_cb & cb) {
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
const int64_t n_embd = hparams.n_embd;
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
struct ggml_tensor * inpL;
if (batch.token) {
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
lctx.inp_tokens = ggml_new_tensor_1d(ctx, GGML_TYPE_I32, batch.n_tokens);
cb(lctx.inp_tokens, "inp_tokens", -1);
ggml_set_input(lctx.inp_tokens);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = ggml_get_rows(ctx, tok_embd, lctx.inp_tokens);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
} else {
#ifdef GGML_USE_MPI
GGML_ASSERT(false && "not implemented");
#endif
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
lctx.inp_embd = ggml_new_tensor_2d(ctx, GGML_TYPE_F32, n_embd, batch.n_tokens);
inpL = lctx.inp_embd;
ggml_set_input(lctx.inp_embd);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
cb(inpL, "inp_embd", -1);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
return inpL;
}
static void llm_build_kv_store(
struct ggml_context * ctx,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
const llama_hparams & hparams,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
const llama_cparams & cparams,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
const llama_kv_cache & kv,
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
struct ggml_cgraph * graph,
struct ggml_tensor * k_cur,
struct ggml_tensor * v_cur,
int32_t n_tokens,
int32_t kv_head,
const llm_build_cb & cb,
int64_t il) {
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
const int64_t n_ctx = cparams.n_ctx;
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_embd_k_gqa = hparams.n_embd_k_gqa();
const int64_t n_embd_v_gqa = hparams.n_embd_v_gqa();
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
GGML_ASSERT(kv.size == n_ctx);
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
struct ggml_tensor * k_cache_view = ggml_view_1d(ctx, kv.k_l[il], n_tokens*n_embd_k_gqa,
(ggml_row_size(kv.k_l[il]->type, n_embd_k_gqa))*kv_head);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
cb(k_cache_view, "k_cache_view", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
// note: storing RoPE-ed version of K in the KV cache
ggml_build_forward_expand(graph, ggml_cpy(ctx, k_cur, k_cache_view));
assert(v_cur->ne[0] == n_embd_v_gqa && v_cur->ne[1] == n_tokens);
struct ggml_tensor * v_cache_view = nullptr;
if (cparams.flash_attn) {
v_cache_view = ggml_view_1d(ctx, kv.v_l[il], n_tokens*n_embd_v_gqa,
(kv_head)*ggml_row_size(kv.v_l[il]->type, n_embd_v_gqa));
} else {
// note: the V cache is transposed when not using flash attention
v_cache_view = ggml_view_2d(ctx, kv.v_l[il], n_tokens, n_embd_v_gqa,
( n_ctx)*ggml_element_size(kv.v_l[il]),
(kv_head)*ggml_element_size(kv.v_l[il]));
v_cur = ggml_transpose(ctx, v_cur);
}
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
cb(v_cache_view, "v_cache_view", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
ggml_build_forward_expand(graph, ggml_cpy(ctx, v_cur, v_cache_view));
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
}
static struct ggml_tensor * llm_build_norm(
struct ggml_context * ctx,
struct ggml_tensor * cur,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
const llama_hparams & hparams,
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
struct ggml_tensor * mw,
struct ggml_tensor * mb,
llm_norm_type type,
const llm_build_cb & cb,
int il) {
switch (type) {
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
case LLM_NORM: cur = ggml_norm (ctx, cur, hparams.f_norm_eps); break;
case LLM_NORM_RMS: cur = ggml_rms_norm(ctx, cur, hparams.f_norm_rms_eps); break;
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
}
if (mw || mb) {
cb(cur, "norm", il);
}
if (mw) {
cur = ggml_mul(ctx, cur, mw);
if (mb) {
cb(cur, "norm_w", il);
}
}
if (mb) {
cur = ggml_add(ctx, cur, mb);
}
return cur;
}
static struct ggml_tensor * llm_build_ffn(
struct ggml_context * ctx,
struct ggml_tensor * cur,
struct ggml_tensor * up,
struct ggml_tensor * up_b,
struct ggml_tensor * gate,
struct ggml_tensor * gate_b,
struct ggml_tensor * down,
struct ggml_tensor * down_b,
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
struct ggml_tensor * act_scales,
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm_ffn_op_type type_op,
llm_ffn_gate_type type_gate,
const llm_build_cb & cb,
int il) {
struct ggml_tensor * tmp = ggml_mul_mat(ctx, up, cur);
cb(tmp, "ffn_up", il);
if (up_b) {
tmp = ggml_add(ctx, tmp, up_b);
cb(tmp, "ffn_up_b", il);
}
if (gate) {
switch (type_gate) {
case LLM_FFN_SEQ:
{
cur = ggml_mul_mat(ctx, gate, tmp);
cb(cur, "ffn_gate", il);
} break;
case LLM_FFN_PAR:
{
cur = ggml_mul_mat(ctx, gate, cur);
cb(cur, "ffn_gate", il);
} break;
}
if (gate_b) {
cur = ggml_add(ctx, cur, gate_b);
cb(cur, "ffn_gate_b", il);
}
} else {
cur = tmp;
}
switch (type_op) {
case LLM_FFN_SILU:
{
cur = ggml_silu(ctx, cur);
cb(cur, "ffn_silu", il);
} break;
case LLM_FFN_GELU:
{
cur = ggml_gelu(ctx, cur);
cb(cur, "ffn_gelu", il);
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
if (act_scales != NULL) {
cur = ggml_div(ctx, cur, act_scales);
cb(cur, "ffn_act", il);
}
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
} break;
case LLM_FFN_RELU:
{
cur = ggml_relu(ctx, cur);
cb(cur, "ffn_relu", il);
} break;
case LLM_FFN_RELU_SQR:
{
cur = ggml_relu(ctx, cur);
cb(cur, "ffn_relu", il);
cur = ggml_sqr(ctx, cur);
cb(cur, "ffn_sqr(relu)", il);
} break;
}
if (type_gate == LLM_FFN_PAR) {
cur = ggml_mul(ctx, cur, tmp);
cb(cur, "ffn_gate_par", il);
}
cur = ggml_mul_mat(ctx, down, cur);
if (down_b) {
cb(cur, "ffn_down", il);
}
if (down_b) {
cur = ggml_add(ctx, cur, down_b);
}
return cur;
}
ggml : group all experts in a single ggml_mul_mat_id (#6505) * ggml : group all experts in a single ggml_mul_mat_id cuda : improve mmid row copy * cuda : fix bin bcast with non-cont src0 * test-backend-ops : only run all mul mat tests for base types * llama : disable moe offloading with SYCL --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-18 13:18:48 +00:00
static struct ggml_tensor * llm_build_moe_ffn(
struct ggml_context * ctx,
struct ggml_tensor * cur,
struct ggml_tensor * gate_inp,
struct ggml_tensor * up_exps,
struct ggml_tensor * gate_exps,
struct ggml_tensor * down_exps,
int64_t n_expert,
int64_t n_expert_used,
llm_ffn_op_type type_op,
bool norm_w,
const llm_build_cb & cb,
int il) {
int64_t n_embd = cur->ne[0];
int64_t n_tokens = cur->ne[1];
ggml_tensor * logits = ggml_mul_mat(ctx, gate_inp, cur); // [n_expert, n_tokens]
cb(logits, "ffn_moe_logits", il);
ggml_tensor * probs = ggml_soft_max(ctx, logits); // [n_expert, n_tokens]
cb(probs, "ffn_moe_probs", il);
// select experts
ggml_tensor * selected_experts = ggml_top_k(ctx, probs, n_expert_used); // [n_expert_used, n_tokens]
cb(selected_experts->src[0], "ffn_moe_argsort", il);
cb(selected_experts, "ffn_moe_topk", il);
ggml_tensor * weights = ggml_get_rows(ctx,
ggml_reshape_3d(ctx, probs, 1, n_expert, n_tokens), selected_experts); // [1, n_expert_used, n_tokens]
cb(weights, "ffn_moe_weights", il);
if (norm_w) {
weights = ggml_reshape_2d(ctx, weights, n_expert_used, n_tokens);
ggml_tensor * weights_sum = ggml_sum_rows(ctx, weights); // [1, n_tokens]
cb(weights_sum, "ffn_moe_weights_sum", il);
weights = ggml_div(ctx, weights, weights_sum); // [n_expert_used, n_tokens]
cb(weights, "ffn_moe_weights_norm", il);
weights = ggml_reshape_3d(ctx, weights, 1, n_expert_used, n_tokens);
}
cur = ggml_reshape_3d(ctx, cur, n_embd, 1, n_tokens);
ggml_tensor * up = ggml_mul_mat_id(ctx, up_exps, cur, selected_experts); // [n_ff, n_expert_used, n_tokens]
cb(up, "ffn_moe_up", il);
ggml_tensor * gate = ggml_mul_mat_id(ctx, gate_exps, cur, selected_experts); // [n_ff, n_expert_used, n_tokens]
cb(gate, "ffn_moe_gate", il);
switch (type_op) {
case LLM_FFN_SILU:
{
gate = ggml_silu(ctx, gate);
cb(gate, "ffn_moe_silu", il);
} break;
case LLM_FFN_GELU:
{
gate = ggml_gelu(ctx, gate);
cb(gate, "ffn_moe_gelu", il);
} break;
default:
GGML_ASSERT(false);
}
ggml_tensor * par = ggml_mul(ctx, up, gate); // [n_ff, n_expert_used, n_tokens]
cb(par, "ffn_moe_gate_par", il);
ggml_tensor * experts = ggml_mul_mat_id(ctx, down_exps, par, selected_experts); // [n_embd, n_expert_used, n_tokens]
cb(experts, "ffn_moe_down", il);
experts = ggml_mul(ctx, experts, weights);
// aggregate experts
ggml_tensor * moe_out = nullptr;
for (int i = 0; i < n_expert_used; ++i) {
ggml_tensor * cur_expert = ggml_view_2d(ctx, experts, n_embd, n_tokens,
experts->nb[2], i*experts->nb[1]);
if (i == 0) {
moe_out = cur_expert;
} else {
moe_out = ggml_add(ctx, moe_out, cur_expert);
}
}
if (n_expert_used == 1) {
// avoid returning a non-contiguous tensor
moe_out = ggml_cont(ctx, moe_out);
}
return moe_out;
}
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
static struct ggml_tensor * llm_build_kqv(
struct ggml_context * ctx,
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
const llama_model & model,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
const llama_hparams & hparams,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
const llama_cparams & cparams,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
const llama_kv_cache & kv,
llama : run all KQV ops on the CPU with no KV offload (#5049) ggml-ci
2024-01-20 15:05:49 +00:00
struct ggml_cgraph * graph,
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
struct ggml_tensor * wo,
struct ggml_tensor * wo_b,
struct ggml_tensor * q_cur,
struct ggml_tensor * kq_mask,
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
struct ggml_tensor * kq_pos,
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
int32_t n_tokens,
int32_t n_kv,
ggml : change ggml_scale to take a float instead of tensor (#4573) * ggml : change ggml_scale to take a float instead of tensor * ggml : fix CPU implementation * tests : fix test-grad0 ggml-ci
2023-12-21 21:20:49 +00:00
float kq_scale,
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
const llm_build_cb & cb,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
int il) {
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
const int64_t n_ctx = cparams.n_ctx;
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_head = hparams.n_head;
const int64_t n_head_kv = hparams.n_head_kv;
const int64_t n_embd_head_k = hparams.n_embd_head_k;
const int64_t n_embd_k_gqa = hparams.n_embd_k_gqa();
const int64_t n_embd_head_v = hparams.n_embd_head_v;
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
struct ggml_tensor * q = ggml_permute(ctx, q_cur, 0, 2, 1, 3);
cb(q, "q", il);
struct ggml_tensor * k =
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
ggml_view_3d(ctx, kv.k_l[il],
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
n_embd_head_k, n_kv, n_head_kv,
ggml_row_size(kv.k_l[il]->type, n_embd_k_gqa),
ggml_row_size(kv.k_l[il]->type, n_embd_head_k),
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
0);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
cb(k, "k", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
struct ggml_tensor * cur;
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
if (cparams.flash_attn) {
GGML_UNUSED(model);
GGML_UNUSED(n_ctx);
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
// note: if this assert triggers, then some check has failed earlier
// the idea is to detect during context creation that ALiBi would be used and disable Flash Attention
GGML_ASSERT(kq_pos == nullptr && "ALiBi is not yet supported with Flash Attention");
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
// split cached v into n_head heads (not transposed)
struct ggml_tensor * v =
ggml_view_3d(ctx, kv.v_l[il],
n_embd_head_v, n_kv, n_head_kv,
ggml_row_size(kv.v_l[il]->type, n_embd_k_gqa),
ggml_row_size(kv.v_l[il]->type, n_embd_head_k),
0);
cb(v, "v", il);
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = ggml_flash_attn_ext(ctx, q, k, v, kq_mask, kq_scale);
if (model.arch == LLM_ARCH_PHI2 || model.arch == LLM_ARCH_PHI3) {
ggml_flash_attn_ext_set_prec(cur, GGML_PREC_F32);
}
cur = ggml_reshape_2d(ctx, cur, n_embd_head_k*n_head, n_tokens);
} else {
struct ggml_tensor * kq = ggml_mul_mat(ctx, k, q);
cb(kq, "kq", il);
if (model.arch == LLM_ARCH_PHI2 || model.arch == LLM_ARCH_PHI3) {
// for this arch, we need to perform the KQ multiplication with F32 precision, otherwise we get NaNs
// ref: https://github.com/ggerganov/llama.cpp/pull/4490#issuecomment-1859055847
ggml_mul_mat_set_prec(kq, GGML_PREC_F32);
}
if (model.arch == LLM_ARCH_GROK) {
// need to do the following:
// multiply by attn_output_multiplyer of 0.08838834764831845
// and then :
// kq = 30 * tanh(kq / 30)
// before the softmax below
//try from phi2
//ggml_mul_mat_set_prec(kq, GGML_PREC_F32);
kq = ggml_tanh(ctx, ggml_scale(ctx, kq, 0.08838834764831845f/30.0f));
kq = ggml_scale(ctx, kq, 30);
}
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
Vulkan Improvements (#5835) * Improve dequant shaders, add fast q4_0 dequant * Optimize dmmv non-kquants for GCN Remove unnecessary SPIR-V shader duplication * Fix q4_0 dequant dispatch sizes Fix backend free bug * Optimize dequant shaders for q4_1, q5_0, q5_1 and q8_0 * Add unary and binary op shader templates * Fix Vulkan check results * Enable non-contiguous support for simple ops * Add argsort Basic q4_0 mmq shader and unit test * Speed up q4_0 dequant code, enable mmq for q4_0 * Rework matmul pipeline selection * Add soft_max alibi support * Add q4_1, q5_0, q5_1 and q8_0 dequant mat mat mul shaders * Add environment variable GGML_VK_FORCE_MAX_ALLOCATION_SIZE to limit max buffer size Rename GGML_VULKAN_DISABLE_F16 to GGML_VK_DISABLE_F16 for consistency
2024-03-05 12:33:42 +00:00
#if defined(GGML_USE_KOMPUTE)
#pragma message("TODO: ALiBi support in ggml_soft_max_ext is not implemented for Kompute")
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
#pragma message(" Falling back to ggml_alibi(). Will become an error in Mar 2024")
#pragma message("ref: https://github.com/ggerganov/llama.cpp/pull/5488")
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
if (hparams.use_alibi) {
kq = ggml_scale(ctx, kq, kq_scale);
cb(kq, "kq_scaled", il);
ggml : add ggml_soft_max_ext (#4256) * metal : implement soft_max_ext * cuda : implement soft_max_ext * ggml : implement soft_max_ext (CPU) * batched-bench : print threads ggml-ci * metal : simplify soft_max encoding ggml-ci * cuda : use 512 threads for soft_max instead of 32 * ggml : update soft max cpu * cuda : do warp-based block reduce * cuda : increase max block size to 1024 * cuda : fix warp reduction initialization of shared mem * metal : warp-based reduction for soft max kernel * metal : warp-based reduce for rms_norm * metal : simplify soft max kernel ggml-ci * alloc : fix build with debug
2023-12-01 08:51:24 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
kq = ggml_alibi(ctx, kq, /*n_past*/ 0, n_head, hparams.f_max_alibi_bias);
cb(kq, "kq_scaled_alibi", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
kq = ggml_add(ctx, kq, kq_mask);
cb(kq, "kq_masked", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
kq = ggml_soft_max(ctx, kq);
cb(kq, "kq_soft_max", il);
} else
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
#endif
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
{
kq = ggml_soft_max_ext(ctx, kq, kq_mask, kq_pos, kq_scale, hparams.f_max_alibi_bias);
cb(kq, "kq_soft_max_ext", il);
}
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
GGML_ASSERT(kv.size == n_ctx);
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
// split cached v into n_head heads
struct ggml_tensor * v =
ggml_view_3d(ctx, kv.v_l[il],
n_kv, n_embd_head_v, n_head_kv,
ggml_element_size(kv.v_l[il])*n_ctx,
ggml_element_size(kv.v_l[il])*n_ctx*n_embd_head_v,
0);
cb(v, "v", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
struct ggml_tensor * kqv = ggml_mul_mat(ctx, v, kq);
cb(kqv, "kqv", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
struct ggml_tensor * kqv_merged = ggml_permute(ctx, kqv, 0, 2, 1, 3);
cb(kqv_merged, "kqv_merged", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = ggml_cont_2d(ctx, kqv_merged, n_embd_head_k*n_head, n_tokens);
cb(cur, "kqv_merged_cont", il);
}
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llama : run all KQV ops on the CPU with no KV offload (#5049) ggml-ci
2024-01-20 15:05:49 +00:00
ggml_build_forward_expand(graph, cur);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
cur = ggml_mul_mat(ctx, wo, cur);
if (wo_b) {
cb(cur, "kqv_wo", il);
}
if (wo_b) {
cur = ggml_add(ctx, cur, wo_b);
}
return cur;
}
llama : run all KQV ops on the CPU with no KV offload (#5049) ggml-ci
2024-01-20 15:05:49 +00:00
static struct ggml_tensor * llm_build_kv(
struct ggml_context * ctx,
const llama_model & model,
const llama_hparams & hparams,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
const llama_cparams & cparams,
llama : run all KQV ops on the CPU with no KV offload (#5049) ggml-ci
2024-01-20 15:05:49 +00:00
const llama_kv_cache & kv,
struct ggml_cgraph * graph,
struct ggml_tensor * wo,
struct ggml_tensor * wo_b,
struct ggml_tensor * k_cur,
struct ggml_tensor * v_cur,
struct ggml_tensor * q_cur,
struct ggml_tensor * kq_mask,
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
struct ggml_tensor * kq_pos,
llama : run all KQV ops on the CPU with no KV offload (#5049) ggml-ci
2024-01-20 15:05:49 +00:00
int32_t n_tokens,
int32_t kv_head,
int32_t n_kv,
float kq_scale,
const llm_build_cb & cb,
int il) {
// these nodes are added to the graph together so that they are not reordered
// by doing so, the number of splits in the graph is reduced
llama : fix not enough space in buffer with Qwen (#5086)
2024-01-22 22:42:41 +00:00
ggml_build_forward_expand(graph, q_cur);
llama : run all KQV ops on the CPU with no KV offload (#5049) ggml-ci
2024-01-20 15:05:49 +00:00
ggml_build_forward_expand(graph, k_cur);
ggml_build_forward_expand(graph, v_cur);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
llm_build_kv_store(ctx, hparams, cparams, kv, graph, k_cur, v_cur, n_tokens, kv_head, cb, il);
llama : run all KQV ops on the CPU with no KV offload (#5049) ggml-ci
2024-01-20 15:05:49 +00:00
struct ggml_tensor * cur;
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kqv(ctx, model, hparams, cparams, kv, graph, wo, wo_b,
q_cur, kq_mask, kq_pos, n_tokens, n_kv, kq_scale, cb, il);
llama : run all KQV ops on the CPU with no KV offload (#5049) ggml-ci
2024-01-20 15:05:49 +00:00
cb(cur, "kqv_out", il);
return cur;
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct llm_build_context {
const llama_model & model;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
llama_context & lctx;
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
const llama_hparams & hparams;
const llama_cparams & cparams;
const llama_batch & batch;
const llama_kv_cache & kv_self;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
const int64_t n_embd;
const int64_t n_layer;
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
const int64_t n_rot;
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
const int64_t n_ctx; // user-specified context size (can be different from n_ctx_train)
const int64_t n_head;
const int64_t n_head_kv;
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_embd_head_k;
const int64_t n_embd_k_gqa;
const int64_t n_embd_head_v;
const int64_t n_embd_v_gqa;
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
const int64_t n_expert;
const int64_t n_expert_used;
llama : add custom RoPE (#2054) * Implement customizable RoPE The original RoPE has pre-defined parameters theta_i = 10000^(−2(i−1)/d), for i in [1, 2, ..., d/2] Our customizable RoPE, ggml_rope_custom_inplace, uses theta_i = scale * base^(−2(i−1)/d), for i in [1, 2, ..., d/2] with the default matches the original scale = 1.0 base = 10000 The new command line arguments --rope-freq-base --rope-freq-scale set the two new RoPE parameter. Recent researches show changing these two parameters extends the context limit with minimal loss. 1. Extending Context to 8K kaiokendev https://kaiokendev.github.io/til#extending-context-to-8k 2. Extending Context Window of Large Language Models via Positional Interpolation Shouyuan Chen, Sherman Wong, Liangjian Chen, Yuandong Tian https://arxiv.org/abs/2306.15595 3. NTK-Aware Scaled RoPE allows LLaMA models to have extended (8k+) context size without any fine-tuning and minimal perplexity degradation. https://www.reddit.com/user/bloc97 https://www.reddit.com/r/LocalLLaMA/comments/14lz7j5/ntkaware_scaled_rope_allows_llama_models_to_have/ For the bold, try adding the following command line parameters to your favorite model: -c 16384 --rope-freq-base 80000 --rope-freq-scale 0.5 * ggml-metal: fix custom rope * common: fix argument names in help * llama: increase MEM_REQ_EVAL for MODEL_3B It avoids crashing for quantized weights on CPU. Better ways to calculate the required buffer size would be better. * llama: make MEM_REQ_EVAL depend on n_ctx * server: use proper Content-Type in curl examples Without the header Content-Type: application/json, curl will POST with Content-Type: application/x-www-form-urlencoded Though our simple server doesn't care, the httplib.h used has a limit with CPPHTTPLIB_FORM_URL_ENCODED_PAYLOAD_MAX_LENGTH 8192 With Content-Type: application/json, we can send large json data. * style : minor fixes, mostly indentations * ggml : fix asserts --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-15 10:34:16 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
const float freq_base;
const float freq_scale;
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
const float ext_factor;
const float attn_factor;
const float beta_fast;
const float beta_slow;
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
const float norm_eps;
const float norm_rms_eps;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
const int32_t n_tokens;
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
const int32_t n_kv; // size of KV cache to consider (n_kv <= kv_self.size)
const int32_t n_outputs;
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
const int32_t kv_head; // index of where we store new KV data in the cache
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
const int32_t n_orig_ctx;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
const bool flash_attn;
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
const enum llama_pooling_type pooling_type;
const enum llama_rope_type rope_type;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
const llm_build_cb & cb;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
std::vector<uint8_t> & buf_compute_meta;
Multi GPU support, CUDA refactor, CUDA scratch buffer (#1703) * CUDA multi GPU + scratch ggml_cuda_compute_forward Tensor parallelism ggml_cuda_add ggml_cuda_rms_norm ggml_cuda_silu CUDA scratch buffer --main-gpu CLI option
2023-06-06 19:33:23 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_context * ctx0 = nullptr;
CUDA full GPU acceleration, KV cache in VRAM (#1827) * Fixed CUDA RoPE * ggml_cuda_mul_mat_vec_p021 * ggml_cuda_scale * ggml_cuda_diag_mask_inf * ggml_is_permuted * ggml_cuda_cpy * flatten rows for ggml_cuda_op * Added a --low-vram option * Fixed Windows performance * Fixed LLAMA_CUDA_DMMV_Y > 1 for WizardLM
2023-06-14 17:47:19 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// TODO: consider making the entire interface noexcept
llm_build_context(
llama_context & lctx,
const llama_batch & batch,
const llm_build_cb & cb,
bool worst_case) :
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
model (lctx.model),
llama : pre-allocate input tensors in a separate buffer (#5100)
2024-01-24 11:48:14 +00:00
lctx (lctx),
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
hparams (model.hparams),
cparams (lctx.cparams),
batch (batch),
kv_self (lctx.kv_self),
n_embd (hparams.n_embd),
n_layer (hparams.n_layer),
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot (hparams.n_rot),
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
n_ctx (cparams.n_ctx),
n_head (hparams.n_head),
n_head_kv (hparams.n_head_kv),
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
n_embd_head_k (hparams.n_embd_head_k),
n_embd_k_gqa (hparams.n_embd_k_gqa()),
n_embd_head_v (hparams.n_embd_head_v),
n_embd_v_gqa (hparams.n_embd_v_gqa()),
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
n_expert (hparams.n_expert),
n_expert_used (hparams.n_expert_used),
freq_base (cparams.rope_freq_base),
freq_scale (cparams.rope_freq_scale),
ext_factor (cparams.yarn_ext_factor),
attn_factor (cparams.yarn_attn_factor),
beta_fast (cparams.yarn_beta_fast),
beta_slow (cparams.yarn_beta_slow),
norm_eps (hparams.f_norm_eps),
norm_rms_eps (hparams.f_norm_rms_eps),
n_tokens (batch.n_tokens),
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
n_kv (worst_case ? kv_self.size : kv_self.n),
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
n_outputs (worst_case ? n_tokens : lctx.n_outputs),
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
kv_head (worst_case ? (kv_self.recurrent ? 0 : kv_self.size - n_tokens) : kv_self.head),
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
n_orig_ctx (cparams.n_yarn_orig_ctx),
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
flash_attn (cparams.flash_attn),
llama : allow for user specified embedding pooling type (#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-03 10:40:27 +00:00
pooling_type (cparams.pooling_type),
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
rope_type (hparams.rope_type),
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
cb (cb),
buf_compute_meta (lctx.buf_compute_meta) {
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// all initializations should be done in init()
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
void init() {
struct ggml_init_params params = {
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
/*.mem_size =*/ buf_compute_meta.size(),
/*.mem_buffer =*/ buf_compute_meta.data(),
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
/*.no_alloc =*/ true,
};
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
ctx0 = ggml_init(params);
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
lctx.inp_tokens = nullptr;
lctx.inp_embd = nullptr;
lctx.inp_pos = nullptr;
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
lctx.inp_out_ids = nullptr;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
lctx.inp_KQ_mask = nullptr;
lctx.inp_KQ_pos = nullptr;
lctx.inp_K_shift = nullptr;
lctx.inp_mean = nullptr;
lctx.inp_cls = nullptr;
lctx.inp_s_copy = nullptr;
lctx.inp_s_mask = nullptr;
lctx.inp_s_seq = nullptr;
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
void free() {
if (ctx0) {
ggml_free(ctx0);
ctx0 = nullptr;
}
}
CUDA full GPU acceleration, KV cache in VRAM (#1827) * Fixed CUDA RoPE * ggml_cuda_mul_mat_vec_p021 * ggml_cuda_scale * ggml_cuda_diag_mask_inf * ggml_is_permuted * ggml_cuda_cpy * flatten rows for ggml_cuda_op * Added a --low-vram option * Fixed Windows performance * Fixed LLAMA_CUDA_DMMV_Y > 1 for WizardLM
2023-06-14 17:47:19 +00:00
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
struct ggml_cgraph * build_k_shift() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
GGML_ASSERT(kv_self.size == n_ctx);
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
lctx.inp_K_shift = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_ctx);
cb(lctx.inp_K_shift, "K_shift", -1);
ggml_set_input(lctx.inp_K_shift);
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * tmp =
// we rotate only the first n_rot dimensions
ggml_rope_custom_inplace(ctx0,
ggml_view_3d(ctx0, kv_self.k_l[il],
n_embd_head_k, n_head_kv, n_ctx,
ggml_row_size(kv_self.k_l[il]->type, n_embd_head_k),
ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa),
0),
lctx.inp_K_shift, n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow);
cb(tmp, "K_shifted", il);
ggml_build_forward_expand(gf, tmp);
}
return gf;
}
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
struct ggml_cgraph * build_s_copy() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
GGML_ASSERT(kv_self.recurrent);
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * state_copy = build_inp_s_copy();
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * conv_states = ggml_reshape_2d(ctx0, kv_self.k_l[il], hparams.n_embd_k_s(), kv_self.size);
struct ggml_tensor * ssm_states = ggml_reshape_2d(ctx0, kv_self.v_l[il], hparams.n_embd_v_s(), kv_self.size);
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
conv_states = ggml_get_rows(ctx0, conv_states, state_copy);
ssm_states = ggml_get_rows(ctx0, ssm_states, state_copy);
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
// TODO: name the intermediate tensors with cb()
ggml_build_forward_expand(gf, ggml_cpy(ctx0, conv_states, kv_self.k_l[il]));
ggml_build_forward_expand(gf, ggml_cpy(ctx0, ssm_states, kv_self.v_l[il]));
}
return gf;
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
struct ggml_cgraph * build_defrag(const std::vector<uint32_t> & ids) {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
for (uint32_t i = 0; i < ids.size(); ++i) {
const uint32_t id = ids[i];
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
if (i == id || id == ids.size()) {
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
continue;
}
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
uint32_t nm = 1;
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
while (i + nm < ids.size() && ids[i + nm] == id + nm) {
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
nm++;
}
for (int il = 0; il < n_layer; ++il) {
ggml_tensor * view_k_src = ggml_view_2d(ctx0, kv_self.k_l[il],
n_embd_k_gqa, nm,
ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa),
ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa*i));
ggml_tensor * view_k_dst = ggml_view_2d(ctx0, kv_self.k_l[il],
n_embd_k_gqa, nm,
ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa),
ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa*id));
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
ggml_tensor * view_v_src;
ggml_tensor * view_v_dst;
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
if (flash_attn) {
// NOTE: the V cache is not transposed when using flash attention
view_v_src = ggml_view_2d(ctx0, kv_self.v_l[il],
n_embd_v_gqa, nm,
ggml_row_size(kv_self.v_l[il]->type, n_embd_v_gqa),
ggml_row_size(kv_self.v_l[il]->type, n_embd_v_gqa*i));
view_v_dst = ggml_view_2d(ctx0, kv_self.v_l[il],
n_embd_v_gqa, nm,
ggml_row_size(kv_self.v_l[il]->type, n_embd_v_gqa),
ggml_row_size(kv_self.v_l[il]->type, n_embd_v_gqa*id));
} else {
view_v_src = ggml_view_2d(ctx0, kv_self.v_l[il],
nm, n_embd_v_gqa,
ggml_row_size(kv_self.v_l[il]->type, kv_self.size),
ggml_row_size(kv_self.v_l[il]->type, i));
view_v_dst = ggml_view_2d(ctx0, kv_self.v_l[il],
nm, n_embd_v_gqa,
ggml_row_size(kv_self.v_l[il]->type, kv_self.size),
ggml_row_size(kv_self.v_l[il]->type, id));
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ggml_build_forward_expand(gf, ggml_cpy(ctx0, view_k_src, view_k_dst));
ggml_build_forward_expand(gf, ggml_cpy(ctx0, view_v_src, view_v_dst));
}
i += nm - 1;
}
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
//LLAMA_LOG_INFO("gf->n_nodes = %d\n", gf->n_nodes);
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
return gf;
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * build_inp_pos() {
lctx.inp_pos = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_tokens);
cb(lctx.inp_pos, "inp_pos", -1);
ggml_set_input(lctx.inp_pos);
return lctx.inp_pos;
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
struct ggml_tensor * build_inp_out_ids() {
lctx.inp_out_ids = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_outputs);
cb(lctx.inp_out_ids, "inp_out_ids", -1);
ggml_set_input(lctx.inp_out_ids);
return lctx.inp_out_ids;
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * build_inp_KQ_mask(bool causal = true) {
if (causal) {
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
lctx.inp_KQ_mask = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_kv, GGML_PAD(n_tokens, GGML_KQ_MASK_PAD));
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
} else {
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
lctx.inp_KQ_mask = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_tokens, GGML_PAD(n_tokens, GGML_KQ_MASK_PAD));
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
}
cb(lctx.inp_KQ_mask, "KQ_mask", -1);
ggml_set_input(lctx.inp_KQ_mask);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
return flash_attn ? ggml_cast(ctx0, lctx.inp_KQ_mask, GGML_TYPE_F16) : lctx.inp_KQ_mask;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
}
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
struct ggml_tensor * build_inp_KQ_pos(bool causal = true) {
if (causal) {
lctx.inp_KQ_pos = ggml_new_tensor_1d(ctx0, GGML_TYPE_F32, n_kv);
} else {
// TODO: this will be needed for ALiBi-based BERT models
// https://github.com/ggerganov/llama.cpp/pull/6826
lctx.inp_KQ_pos = ggml_new_tensor_1d(ctx0, GGML_TYPE_F32, n_tokens);
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
cb(lctx.inp_KQ_pos, "KQ_pos", -1);
ggml_set_input(lctx.inp_KQ_pos);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
return flash_attn ? ggml_cast(ctx0, lctx.inp_KQ_pos, GGML_TYPE_F16) : lctx.inp_KQ_pos;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
}
struct ggml_tensor * build_inp_mean() {
lctx.inp_mean = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_tokens, n_tokens);
cb(lctx.inp_mean, "inp_mean", -1);
ggml_set_input(lctx.inp_mean);
return lctx.inp_mean;
}
struct ggml_tensor * build_inp_cls() {
lctx.inp_cls = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, n_tokens);
cb(lctx.inp_cls, "inp_cls", -1);
ggml_set_input(lctx.inp_cls);
return lctx.inp_cls;
}
struct ggml_tensor * build_inp_s_copy() {
lctx.inp_s_copy = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, kv_self.size);
cb(lctx.inp_s_copy, "inp_s_copy", -1);
ggml_set_input(lctx.inp_s_copy);
return lctx.inp_s_copy;
}
struct ggml_tensor * build_inp_s_mask() {
lctx.inp_s_mask = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, 1, n_kv);
cb(lctx.inp_s_mask, "inp_s_mask", -1);
ggml_set_input(lctx.inp_s_mask);
return lctx.inp_s_mask;
}
struct ggml_tensor * build_inp_s_seq() {
lctx.inp_s_seq = ggml_new_tensor_2d(ctx0, GGML_TYPE_I32, n_kv, n_tokens);
cb(lctx.inp_s_seq, "inp_s_seq", -1);
ggml_set_input(lctx.inp_s_seq);
return lctx.inp_s_seq;
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_cgraph * build_llama() {
sync : ggml (backend v2) (#3912) * sync : ggml (backend v2) (wip) * sync : migrate examples and llama.cpp to dynamic graphs (wip) * sync : update tests + fix max op params to 64 ggml-ci * sync : ggml-cuda ggml-ci * llama : fix save/load state context size ggml-ci * sync : try to fix build on tvOS * sync : pass custom graph sizes in training examples * sync : update graph copies to new ggml API * sync : update sync-ggml.sh with new files * scripts : fix header in sync script * train : fix context size calculations * llama : increase inference graph size up to 4096 nodes * train : allocate grads for backward graphs * train : allocate grads for gb_tmp
2023-11-13 12:16:23 +00:00
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
ggml, llama : avoid heavy V transpose + improvements (#775) ggml : - added ggml_view_3d() - ggml_view_tensor() now inherits the stride too - reimplement ggml_cpy() to account for dst stride - no longer require tensor->data to be memory aligned llama : - compute RoPE on 32-bit tensors (should be more accurate) - store RoPE-ed K in the KV cache - store transposed V in the KV cache (significant speed-up) - avoid unnecessary Q copy
2023-04-05 19:07:33 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
// mutable variable, needed during the last layer of the computation to skip unused tokens
int32_t n_tokens = this->n_tokens;
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_embd_head = hparams.n_embd_head_v;
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
GGML_ASSERT(n_embd_head == hparams.n_rot);
CUDA full GPU acceleration, KV cache in VRAM (#1827) * Fixed CUDA RoPE * ggml_cuda_mul_mat_vec_p021 * ggml_cuda_scale * ggml_cuda_diag_mask_inf * ggml_is_permuted * ggml_cuda_cpy * flatten rows for ggml_cuda_op * Added a --low-vram option * Fixed Windows performance * Fixed LLAMA_CUDA_DMMV_Y > 1 for WizardLM
2023-06-14 17:47:19 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// inp_pos - contains the positions
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * inp_pos = build_inp_pos();
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * inpSA = inpL;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// norm
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "attn_norm", il);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// self-attention
{
// compute Q and K and RoPE them
struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
cb(Qcur, "Qcur", il);
llama : support attention bias on LLaMA architecture (#4283) * Support attention_bias on LLaMA architecture QKVO bias, should fix InternLM (https://github.com/ggerganov/llama.cpp/issues/3133) and works for LLaMAfied Qwen models (https://github.com/ggerganov/llama.cpp/pull/3743#issuecomment-1825923608). * check existence of qkvo bias while loading llama models Tested on LLaMA2, CUDA and CPU. * Update llama.cpp
2023-12-01 18:17:06 +00:00
if (model.layers[il].bq) {
Qcur = ggml_add(ctx0, Qcur, model.layers[il].bq);
cb(Qcur, "Qcur", il);
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur);
cb(Kcur, "Kcur", il);
llama : support attention bias on LLaMA architecture (#4283) * Support attention_bias on LLaMA architecture QKVO bias, should fix InternLM (https://github.com/ggerganov/llama.cpp/issues/3133) and works for LLaMAfied Qwen models (https://github.com/ggerganov/llama.cpp/pull/3743#issuecomment-1825923608). * check existence of qkvo bias while loading llama models Tested on LLaMA2, CUDA and CPU. * Update llama.cpp
2023-12-01 18:17:06 +00:00
if (model.layers[il].bk) {
Kcur = ggml_add(ctx0, Kcur, model.layers[il].bk);
cb(Kcur, "Kcur", il);
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur);
cb(Vcur, "Vcur", il);
llama : support attention bias on LLaMA architecture (#4283) * Support attention_bias on LLaMA architecture QKVO bias, should fix InternLM (https://github.com/ggerganov/llama.cpp/issues/3133) and works for LLaMAfied Qwen models (https://github.com/ggerganov/llama.cpp/pull/3743#issuecomment-1825923608). * check existence of qkvo bias while loading llama models Tested on LLaMA2, CUDA and CPU. * Update llama.cpp
2023-12-01 18:17:06 +00:00
if (model.layers[il].bv) {
Vcur = ggml_add(ctx0, Vcur, model.layers[il].bv);
cb(Vcur, "Vcur", il);
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
Qcur = ggml_rope_custom(
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cb(Qcur, "Qcur", il);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
Kcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cb(Kcur, "Kcur", il);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llama : support attention bias on LLaMA architecture (#4283) * Support attention_bias on LLaMA architecture QKVO bias, should fix InternLM (https://github.com/ggerganov/llama.cpp/issues/3133) and works for LLaMAfied Qwen models (https://github.com/ggerganov/llama.cpp/pull/3743#issuecomment-1825923608). * check existence of qkvo bias while loading llama models Tested on LLaMA2, CUDA and CPU. * Update llama.cpp
2023-12-01 18:17:06 +00:00
model.layers[il].wo, model.layers[il].bo,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
}
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
n_tokens = n_outputs;
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
cb(ffn_inp, "ffn_inp", il);
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// feed-forward network
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
if (model.layers[il].ffn_gate_inp == nullptr) {
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "ffn_norm", il);
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, NULL,
model.layers[il].ffn_gate, NULL,
model.layers[il].ffn_down, NULL,
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
NULL,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
cb(cur, "ffn_out", il);
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
} else {
// MoE branch
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "ffn_norm", il);
ggml : group all experts in a single ggml_mul_mat_id (#6505) * ggml : group all experts in a single ggml_mul_mat_id cuda : improve mmid row copy * cuda : fix bin bcast with non-cont src0 * test-backend-ops : only run all mul mat tests for base types * llama : disable moe offloading with SYCL --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-18 13:18:48 +00:00
cur = llm_build_moe_ffn(ctx0, cur,
model.layers[il].ffn_gate_inp,
model.layers[il].ffn_up_exps,
model.layers[il].ffn_gate_exps,
model.layers[il].ffn_down_exps,
n_expert, n_expert_used,
LLM_FFN_SILU, true,
cb, il);
cb(cur, "ffn_moe_out", il);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
}
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = ggml_add(ctx0, cur, ffn_inp);
llama : add support for control vectors (#5970) * control vector api and implementation * control-vectors : minor code style updates * disable control vector when data == nullptr use -1 for disabled range (also on init) in case we ever support controlling layer 0 (embeddings) --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-15 20:43:02 +00:00
cb(cur, "ffn_out", il);
ggml_tensor * layer_dir = lctx.cvec.tensor_for(il);
if (layer_dir != nullptr) {
cur = ggml_add(ctx0, cur, layer_dir);
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cb(cur, "l_out", il);
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// input for next layer
inpL = cur;
}
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = inpL;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm, NULL,
LLM_NORM_RMS, cb, -1);
cb(cur, "result_norm", -1);
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// lm_head
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
ggml_build_forward_expand(gf, cur);
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
return gf;
}
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_cgraph * build_baichuan() {
sync : ggml (backend v2) (#3912) * sync : ggml (backend v2) (wip) * sync : migrate examples and llama.cpp to dynamic graphs (wip) * sync : update tests + fix max op params to 64 ggml-ci * sync : ggml-cuda ggml-ci * llama : fix save/load state context size ggml-ci * sync : try to fix build on tvOS * sync : pass custom graph sizes in training examples * sync : update graph copies to new ggml API * sync : update sync-ggml.sh with new files * scripts : fix header in sync script * train : fix context size calculations * llama : increase inference graph size up to 4096 nodes * train : allocate grads for backward graphs * train : allocate grads for gb_tmp
2023-11-13 12:16:23 +00:00
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_embd_head = hparams.n_embd_head_v;
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
llama : fix llm_build_k_shift to use correct n_rot (#4889) * llama : fix llm_build_k_shift to use correct n_rot ggml-ci * llama : always use hparams.n_rot for ggml_rope_custom ggml-ci * convert : fix persimmon conversion to write correct n_rot
2024-01-12 11:01:56 +00:00
GGML_ASSERT(n_embd_head == hparams.n_rot);
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// inp_pos - contains the positions
llama : fix Baichuan2 13B (#6092)
2024-03-15 21:14:16 +00:00
struct ggml_tensor * inp_pos = model.type == MODEL_7B ? build_inp_pos() : nullptr;
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
// positions of the tokens in the KV cache
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_pos = build_inp_KQ_pos();
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * inpSA = inpL;
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "attn_norm", il);
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// self-attention
{
struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
cb(Qcur, "Qcur", il);
struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur);
cb(Kcur, "Kcur", il);
struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur);
cb(Vcur, "Vcur", il);
switch (model.type) {
case MODEL_7B:
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
Qcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
);
Kcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
break;
case MODEL_13B:
Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd/n_head, n_head, n_tokens);
Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd/n_head, n_head, n_tokens);
break;
default:
GGML_ASSERT(false);
}
cb(Qcur, "Qcur", il);
cb(Kcur, "Kcur", il);
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
model.layers[il].wo, NULL,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, KQ_pos, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
cb(ffn_inp, "ffn_inp", il);
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// feed-forward network
{
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "ffn_norm", il);
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, NULL,
model.layers[il].ffn_gate, NULL,
model.layers[il].ffn_down, NULL,
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
NULL,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
cb(cur, "ffn_out", il);
}
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "l_out", il);
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// input for next layer
inpL = cur;
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = inpL;
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm, NULL,
LLM_NORM_RMS, cb, -1);
cb(cur, "result_norm", -1);
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// lm_head
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
ggml_build_forward_expand(gf, cur);
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
return gf;
}
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
[Model] Add support for xverse (#6301) * Support xverse model convert to gguf format. * 1. Convert xverse models to gguf; 2. Add LLM_ARCH_XVERSE inference in llama.cpp; 3. Add xverse item in Supported models in README.md; * * gguf-py: remove redundant logs * llama: remove the init_mapping_prefetch custom parameter * llama.cpp: Include the changes from #6122 to exclude the unused outputs of the last layers. * - Fix format issues - Remove duplicate set kqv_out to llm_build_kv * Update llama.cpp --------- Co-authored-by: willhe <willhe@xverse.cn> Co-authored-by: willhe <hexin@xverse.cn>
2024-03-29 13:37:03 +00:00
struct ggml_cgraph * build_xverse() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
const int64_t n_embd_head = hparams.n_embd_head_v;
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
GGML_ASSERT(n_embd_head == hparams.n_rot);
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
// inp_pos - contains the positions
struct ggml_tensor * inp_pos = build_inp_pos();
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
// positions of the tokens in the KV cache
struct ggml_tensor * KQ_pos = build_inp_KQ_pos();
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * inpSA = inpL;
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "attn_norm", il);
// self-attention
{
struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
cb(Qcur, "Qcur", il);
struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur);
cb(Kcur, "Kcur", il);
struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur);
cb(Vcur, "Vcur", il);
Qcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens), inp_pos,
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Qcur, "Qcur", il);
Kcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens), inp_pos,
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
[Model] Add support for xverse (#6301) * Support xverse model convert to gguf format. * 1. Convert xverse models to gguf; 2. Add LLM_ARCH_XVERSE inference in llama.cpp; 3. Add xverse item in Supported models in README.md; * * gguf-py: remove redundant logs * llama: remove the init_mapping_prefetch custom parameter * llama.cpp: Include the changes from #6122 to exclude the unused outputs of the last layers. * - Fix format issues - Remove duplicate set kqv_out to llm_build_kv * Update llama.cpp --------- Co-authored-by: willhe <willhe@xverse.cn> Co-authored-by: willhe <hexin@xverse.cn>
2024-03-29 13:37:03 +00:00
model.layers[il].wo, NULL,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, KQ_pos, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
[Model] Add support for xverse (#6301) * Support xverse model convert to gguf format. * 1. Convert xverse models to gguf; 2. Add LLM_ARCH_XVERSE inference in llama.cpp; 3. Add xverse item in Supported models in README.md; * * gguf-py: remove redundant logs * llama: remove the init_mapping_prefetch custom parameter * llama.cpp: Include the changes from #6122 to exclude the unused outputs of the last layers. * - Fix format issues - Remove duplicate set kqv_out to llm_build_kv * Update llama.cpp --------- Co-authored-by: willhe <willhe@xverse.cn> Co-authored-by: willhe <hexin@xverse.cn>
2024-03-29 13:37:03 +00:00
}
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
}
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
cb(ffn_inp, "ffn_inp", il);
// feed-forward network
{
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "ffn_norm", il);
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, NULL,
model.layers[il].ffn_gate, NULL,
model.layers[il].ffn_down, NULL,
NULL,
LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
cb(cur, "ffn_out", il);
}
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "l_out", il);
// input for next layer
inpL = cur;
}
cur = inpL;
cur = llm_build_norm(ctx0, cur, hparams, model.output_norm, NULL, LLM_NORM_RMS, cb, -1);
cb(cur, "result_norm", -1);
// lm_head
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
return gf;
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_cgraph * build_falcon() {
sync : ggml (backend v2) (#3912) * sync : ggml (backend v2) (wip) * sync : migrate examples and llama.cpp to dynamic graphs (wip) * sync : update tests + fix max op params to 64 ggml-ci * sync : ggml-cuda ggml-ci * llama : fix save/load state context size ggml-ci * sync : try to fix build on tvOS * sync : pass custom graph sizes in training examples * sync : update graph copies to new ggml API * sync : update sync-ggml.sh with new files * scripts : fix header in sync script * train : fix context size calculations * llama : increase inference graph size up to 4096 nodes * train : allocate grads for backward graphs * train : allocate grads for gb_tmp
2023-11-13 12:16:23 +00:00
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_embd_head = hparams.n_embd_head_v;
const int64_t n_embd_gqa = hparams.n_embd_v_gqa();
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
llama : fix llm_build_k_shift to use correct n_rot (#4889) * llama : fix llm_build_k_shift to use correct n_rot ggml-ci * llama : always use hparams.n_rot for ggml_rope_custom ggml-ci * convert : fix persimmon conversion to write correct n_rot
2024-01-12 11:01:56 +00:00
GGML_ASSERT(n_embd_head == hparams.n_rot);
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// inp_pos - contains the positions
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * inp_pos = build_inp_pos();
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * attn_norm;
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
attn_norm = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm,
model.layers[il].attn_norm_b,
LLM_NORM, cb, il);
cb(attn_norm, "attn_norm", il);
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// self-attention
{
if (model.layers[il].attn_norm_2) {
// Falcon-40B
llm : fix falcon norm after refactoring (#3837)
2023-11-01 21:00:50 +00:00
cur = llm_build_norm(ctx0, inpL, hparams,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
model.layers[il].attn_norm_2,
model.layers[il].attn_norm_2_b,
LLM_NORM, cb, il);
cb(cur, "attn_norm_2", il);
} else {
cur = attn_norm;
}
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = ggml_mul_mat(ctx0, model.layers[il].wqkv, cur);
cb(cur, "wqkv", il);
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * Qcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd, n_tokens, cur->nb[1], 0*sizeof(float)*(n_embd)));
struct ggml_tensor * Kcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd)));
struct ggml_tensor * Vcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd + n_embd_gqa)));
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cb(Qcur, "Qcur", il);
cb(Kcur, "Kcur", il);
cb(Vcur, "Vcur", il);
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// using mode = 2 for neox mode
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
Qcur = ggml_rope_custom(
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ctx0, Qcur, inp_pos, n_rot, rope_type, 0, n_orig_ctx,
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
freq_base, freq_scale, ext_factor, attn_factor, beta_fast, beta_slow
);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cb(Qcur, "Qcur", il);
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
Kcur = ggml_rope_custom(
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ctx0, Kcur, inp_pos, n_rot, rope_type, 0, n_orig_ctx,
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
freq_base, freq_scale, ext_factor, attn_factor, beta_fast, beta_slow
);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cb(Kcur, "Kcur", il);
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
model.layers[il].wo, NULL,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
}
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
attn_norm = ggml_get_rows(ctx0, attn_norm, inp_out_ids);
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * ffn_inp = cur;
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// feed forward
{
cur = llm_build_ffn(ctx0, attn_norm, // !! use the attn norm, not the result
model.layers[il].ffn_up, NULL,
NULL, NULL,
model.layers[il].ffn_down, NULL,
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
NULL,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
LLM_FFN_GELU, LLM_FFN_SEQ, cb, il);
cb(cur, "ffn_out", il);
}
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "l_out", il);
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = ggml_add(ctx0, cur, inpL);
cb(cur, "l_out", il);
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// input for next layer
inpL = cur;
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = inpL;
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// norm
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm,
model.output_norm_b,
LLM_NORM, cb, -1);
cb(cur, "result_norm", -1);
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
ggml_build_forward_expand(gf, cur);
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
return gf;
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
}
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
struct ggml_cgraph * build_grok() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
// mutable variable, needed during the last layer of the computation to skip unused tokens
int32_t n_tokens = this->n_tokens;
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
const int64_t n_embd_head = hparams.n_embd_head_v;
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
GGML_ASSERT(n_embd_head == hparams.n_rot);
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
// multiply by embedding_multiplier_scale of 78.38367176906169
inpL = ggml_scale(ctx0, inpL, 78.38367176906169f);
// inp_pos - contains the positions
struct ggml_tensor * inp_pos = build_inp_pos();
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * inpSA = inpL;
// norm
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "attn_norm", il);
// self-attention
{
// compute Q and K and RoPE them
struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
cb(Qcur, "Qcur", il);
if (model.layers[il].bq) {
Qcur = ggml_add(ctx0, Qcur, model.layers[il].bq);
cb(Qcur, "Qcur", il);
}
struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur);
cb(Kcur, "Kcur", il);
if (model.layers[il].bk) {
Kcur = ggml_add(ctx0, Kcur, model.layers[il].bk);
cb(Kcur, "Kcur", il);
}
struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur);
cb(Vcur, "Vcur", il);
if (model.layers[il].bv) {
Vcur = ggml_add(ctx0, Vcur, model.layers[il].bv);
cb(Vcur, "Vcur", il);
}
Qcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens), inp_pos,
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Qcur, "Qcur", il);
Kcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens), inp_pos,
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model.layers[il].wo, model.layers[il].bo,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f, cb, il);
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
n_tokens = n_outputs;
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
}
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
// Grok
// if attn_out_norm is present then apply it before adding the input
if (model.layers[il].attn_out_norm) {
cur = llm_build_norm(ctx0, cur, hparams,
model.layers[il].attn_out_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "attn_out_norm", il);
}
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
cb(ffn_inp, "ffn_inp", il);
// feed-forward network
// MoE branch
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "ffn_norm", il);
ggml : group all experts in a single ggml_mul_mat_id (#6505) * ggml : group all experts in a single ggml_mul_mat_id cuda : improve mmid row copy * cuda : fix bin bcast with non-cont src0 * test-backend-ops : only run all mul mat tests for base types * llama : disable moe offloading with SYCL --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-18 13:18:48 +00:00
cur = llm_build_moe_ffn(ctx0, cur,
model.layers[il].ffn_gate_inp,
model.layers[il].ffn_up_exps,
model.layers[il].ffn_gate_exps,
model.layers[il].ffn_down_exps,
n_expert, n_expert_used,
LLM_FFN_GELU, true,
cb, il);
cb(cur, "ffn_moe_out", il);
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
// Grok
// if layer_out_norm is present then apply it before adding the input
// Idea: maybe ffn_out_norm is a better name
if (model.layers[il].layer_out_norm) {
cur = llm_build_norm(ctx0, cur, hparams,
model.layers[il].layer_out_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "layer_out_norm", il);
}
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "ffn_out", il);
ggml_tensor * layer_dir = lctx.cvec.tensor_for(il);
if (layer_dir != nullptr) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);
// input for next layer
inpL = cur;
}
cur = inpL;
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm, NULL,
LLM_NORM_RMS, cb, -1);
cb(cur, "result_norm", -1);
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
// lm_head
cur = ggml_mul_mat(ctx0, model.output, cur);
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
// Grok
// multiply logits by output_multiplier_scale of 0.5773502691896257
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
cur = ggml_scale(ctx0, cur, 0.5773502691896257f);
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
cb(cur, "result_output", -1);
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
ggml_build_forward_expand(gf, cur);
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
return gf;
}
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
struct ggml_cgraph * build_dbrx() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
// mutable variable, needed during the last layer of the computation to skip unused tokens
int32_t n_tokens = this->n_tokens;
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
const int64_t n_embd_head = hparams.n_embd_head_v;
const int64_t n_embd_gqa = hparams.n_embd_v_gqa();
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
GGML_ASSERT(n_embd_head == hparams.n_rot);
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
// inp_pos - contains the positions
struct ggml_tensor * inp_pos = build_inp_pos();
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * inpSA = inpL;
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
// norm
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm, NULL,
LLM_NORM, cb, il);
cb(cur, "attn_norm", il);
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
// self-attention
{
struct ggml_tensor * Qcur = nullptr;
struct ggml_tensor * Kcur = nullptr;
struct ggml_tensor * Vcur = nullptr;
cur = ggml_mul_mat(ctx0, model.layers[il].wqkv, cur);
cb(cur, "wqkv", il);
cur = ggml_clamp(ctx0, cur, -hparams.f_clamp_kqv, hparams.f_clamp_kqv);
cb(cur, "wqkv_clamped", il);
Qcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd, n_tokens, cur->nb[1], 0*sizeof(float)*(n_embd)));
Kcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd)));
Vcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd + n_embd_gqa)));
cb(Qcur, "Qcur", il);
cb(Kcur, "Kcur", il);
cb(Vcur, "Vcur", il);
Qcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens), inp_pos,
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Qcur, "Qcur", il);
Kcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens), inp_pos,
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
model.layers[il].wo, NULL,
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
}
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
n_tokens = n_outputs;
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
}
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
cb(ffn_inp, "ffn_inp", il);
// feed-forward network
// MoE branch
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].attn_out_norm, NULL,
LLM_NORM, cb, il);
cb(cur, "attn_out_norm", il);
ggml : group all experts in a single ggml_mul_mat_id (#6505) * ggml : group all experts in a single ggml_mul_mat_id cuda : improve mmid row copy * cuda : fix bin bcast with non-cont src0 * test-backend-ops : only run all mul mat tests for base types * llama : disable moe offloading with SYCL --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-18 13:18:48 +00:00
cur = llm_build_moe_ffn(ctx0, cur,
model.layers[il].ffn_gate_inp,
model.layers[il].ffn_up_exps,
model.layers[il].ffn_gate_exps,
model.layers[il].ffn_down_exps,
n_expert, n_expert_used,
LLM_FFN_SILU, true,
cb, il);
cb(cur, "ffn_moe_out", il);
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "ffn_out", il);
ggml_tensor * layer_dir = lctx.cvec.tensor_for(il);
if (layer_dir != nullptr) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);
// input for next layer
inpL = cur;
}
cur = inpL;
cur = llm_build_norm(ctx0, cur, hparams,
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
model.output_norm, NULL,
LLM_NORM, cb, -1);
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
cb(cur, "result_norm", -1);
// lm_head
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
return gf;
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_cgraph * build_starcoder() {
sync : ggml (backend v2) (#3912) * sync : ggml (backend v2) (wip) * sync : migrate examples and llama.cpp to dynamic graphs (wip) * sync : update tests + fix max op params to 64 ggml-ci * sync : ggml-cuda ggml-ci * llama : fix save/load state context size ggml-ci * sync : try to fix build on tvOS * sync : pass custom graph sizes in training examples * sync : update graph copies to new ggml API * sync : update sync-ggml.sh with new files * scripts : fix header in sync script * train : fix context size calculations * llama : increase inference graph size up to 4096 nodes * train : allocate grads for backward graphs * train : allocate grads for gb_tmp
2023-11-13 12:16:23 +00:00
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_embd_head = hparams.n_embd_head_v;
const int64_t n_embd_gqa = hparams.n_embd_v_gqa();
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// inp_pos - contains the positions
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * inp_pos = build_inp_pos();
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * pos = ggml_get_rows(ctx0, model.pos_embd, inp_pos);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cb(pos, "pos_embd", -1);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
inpL = ggml_add(ctx0, inpL, pos);
cb(inpL, "inpL", -1);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
for (int il = 0; il < n_layer; ++il) {
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm,
model.layers[il].attn_norm_b,
LLM_NORM, cb, il);
cb(cur, "attn_norm", il);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// self-attention
{
cur = ggml_mul_mat(ctx0, model.layers[il].wqkv, cur);
cb(cur, "wqkv", il);
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = ggml_add(ctx0, cur, model.layers[il].bqkv);
cb(cur, "bqkv", il);
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * Qcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd, n_tokens, cur->nb[1], 0*sizeof(float)*(n_embd)));
struct ggml_tensor * Kcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd)));
struct ggml_tensor * Vcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd + n_embd_gqa)));
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cb(Qcur, "Qcur", il);
cb(Kcur, "Kcur", il);
cb(Vcur, "Vcur", il);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
model.layers[il].wo, model.layers[il].bo,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// add the input
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL);
cb(ffn_inp, "ffn_inp", il);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// FF
{
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm,
model.layers[il].ffn_norm_b,
LLM_NORM, cb, il);
cb(cur, "ffn_norm", il);
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, model.layers[il].ffn_up_b,
NULL, NULL,
model.layers[il].ffn_down, model.layers[il].ffn_down_b,
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
NULL,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
LLM_FFN_GELU, LLM_FFN_SEQ, cb, il);
cb(cur, "ffn_out", il);
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
inpL = ggml_add(ctx0, cur, ffn_inp);
cb(inpL, "l_out", il);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = llm_build_norm(ctx0, inpL, hparams,
model.output_norm,
model.output_norm_b,
LLM_NORM, cb, -1);
cb(cur, "result_norm", -1);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
return gf;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_cgraph * build_persimmon() {
sync : ggml (backend v2) (#3912) * sync : ggml (backend v2) (wip) * sync : migrate examples and llama.cpp to dynamic graphs (wip) * sync : update tests + fix max op params to 64 ggml-ci * sync : ggml-cuda ggml-ci * llama : fix save/load state context size ggml-ci * sync : try to fix build on tvOS * sync : pass custom graph sizes in training examples * sync : update graph copies to new ggml API * sync : update sync-ggml.sh with new files * scripts : fix header in sync script * train : fix context size calculations * llama : increase inference graph size up to 4096 nodes * train : allocate grads for backward graphs * train : allocate grads for gb_tmp
2023-11-13 12:16:23 +00:00
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_embd_head = hparams.n_embd_head_v;
llama : fix llm_build_k_shift to use correct n_rot (#4889) * llama : fix llm_build_k_shift to use correct n_rot ggml-ci * llama : always use hparams.n_rot for ggml_rope_custom ggml-ci * convert : fix persimmon conversion to write correct n_rot
2024-01-12 11:01:56 +00:00
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
GGML_ASSERT(n_embd_head/2 == hparams.n_rot);
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
// inp_pos - contains the positions
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * inp_pos = build_inp_pos();
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * residual = inpL;
starcoder : add GPU offloading (#3827) * starcoder : do not GPU split 1D bias tensors * starcoder : offload layers to GPU ggml-ci
2023-10-28 09:06:08 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm,
model.layers[il].attn_norm_b,
LLM_NORM, cb, il);
cb(cur, "attn_norm", il);
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// self attention
{
cur = ggml_mul_mat(ctx0, model.layers[il].wqkv, cur);
cb(cur, "wqkv", il);
cur = ggml_add(ctx0, cur, model.layers[il].bqkv);
cb(cur, "bqkv", il);
// split qkv
GGML_ASSERT(n_head_kv == n_head);
struct ggml_tensor * tmpqkv = ggml_reshape_4d(ctx0, cur, n_embd_head, 3, n_head, n_tokens);
cb(tmpqkv, "tmpqkv", il);
struct ggml_tensor * tmpqkv_perm = ggml_cont(ctx0, ggml_permute(ctx0, tmpqkv, 0, 3, 1, 2));
cb(tmpqkv_perm, "tmpqkv", il);
struct ggml_tensor * tmpq = ggml_view_3d(
ctx0, tmpqkv_perm, n_embd_head, n_head, n_tokens,
ggml_element_size(tmpqkv_perm) * n_embd_head,
ggml_element_size(tmpqkv_perm) * n_embd_head * n_head,
0
);
cb(tmpq, "tmpq", il);
struct ggml_tensor * tmpk = ggml_view_3d(
ctx0, tmpqkv_perm, n_embd_head, n_head, n_tokens,
ggml_element_size(tmpqkv_perm) * n_embd_head,
ggml_element_size(tmpqkv_perm) * n_embd_head * n_head,
ggml_element_size(tmpqkv_perm) * n_embd_head * n_head * n_tokens
);
cb(tmpk, "tmpk", il);
// Q/K Layernorm
tmpq = llm_build_norm(ctx0, tmpq, hparams,
model.layers[il].attn_q_norm,
model.layers[il].attn_q_norm_b,
LLM_NORM, cb, il);
cb(tmpq, "tmpq", il);
tmpk = llm_build_norm(ctx0, tmpk, hparams,
model.layers[il].attn_k_norm,
model.layers[il].attn_k_norm_b,
LLM_NORM, cb, il);
cb(tmpk, "tmpk", il);
// RoPE the first n_rot of q/k, pass the other half, and concat.
struct ggml_tensor * qrot = ggml_view_3d(
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ctx0, tmpq, n_rot, n_head, n_tokens,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
ggml_element_size(tmpq) * n_embd_head,
ggml_element_size(tmpq) * n_embd_head * n_head,
0
);
cb(qrot, "qrot", il);
struct ggml_tensor * krot = ggml_view_3d(
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ctx0, tmpk, n_rot, n_head, n_tokens,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
ggml_element_size(tmpk) * n_embd_head,
ggml_element_size(tmpk) * n_embd_head * n_head,
0
);
cb(krot, "krot", il);
// get the second half of tmpq, e.g tmpq[n_rot:, :, :]
struct ggml_tensor * qpass = ggml_view_3d(
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ctx0, tmpq, n_rot, n_head, n_tokens,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
ggml_element_size(tmpq) * n_embd_head,
ggml_element_size(tmpq) * n_embd_head * n_head,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ggml_element_size(tmpq) * n_rot
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
);
cb(qpass, "qpass", il);
struct ggml_tensor * kpass = ggml_view_3d(
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ctx0, tmpk, n_rot, n_head, n_tokens,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
ggml_element_size(tmpk) * n_embd_head,
ggml_element_size(tmpk) * n_embd_head * n_head,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ggml_element_size(tmpk) * n_rot
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
);
cb(kpass, "kpass", il);
struct ggml_tensor * qrotated = ggml_rope_custom(
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ctx0, qrot, inp_pos, n_rot, rope_type, 0, n_orig_ctx,
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
freq_base, freq_scale, ext_factor, attn_factor, beta_fast, beta_slow
);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cb(qrotated, "qrotated", il);
struct ggml_tensor * krotated = ggml_rope_custom(
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ctx0, krot, inp_pos, n_rot, rope_type, 0, n_orig_ctx,
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
freq_base, freq_scale, ext_factor, attn_factor, beta_fast, beta_slow
);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cb(krotated, "krotated", il);
// ggml currently only supports concatenation on dim=2
// so we need to permute qrot, qpass, concat, then permute back.
qrotated = ggml_cont(ctx0, ggml_permute(ctx0, qrotated, 2, 1, 0, 3));
cb(qrotated, "qrotated", il);
krotated = ggml_cont(ctx0, ggml_permute(ctx0, krotated, 2, 1, 0, 3));
cb(krotated, "krotated", il);
qpass = ggml_cont(ctx0, ggml_permute(ctx0, qpass, 2, 1, 0, 3));
cb(qpass, "qpass", il);
kpass = ggml_cont(ctx0, ggml_permute(ctx0, kpass, 2, 1, 0, 3));
cb(kpass, "kpass", il);
struct ggml_tensor * Qcur = ggml_concat(ctx0, qrotated, qpass);
cb(Qcur, "Qcur", il);
struct ggml_tensor * Kcur = ggml_concat(ctx0, krotated, kpass);
cb(Kcur, "Kcur", il);
Unbreak persimmon after #3837 (#4010)
2023-11-10 13:24:54 +00:00
struct ggml_tensor * Q = ggml_cont(ctx0, ggml_permute(ctx0, Qcur, 2, 1, 0, 3));
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cb(Q, "Q", il);
Kcur = ggml_cont(ctx0, ggml_permute(ctx0, Kcur, 2, 1, 0, 3));
cb(Kcur, "Kcur", il);
struct ggml_tensor * Vcur = ggml_view_3d(
ctx0, tmpqkv_perm, n_embd_head, n_head, n_tokens,
ggml_element_size(tmpqkv_perm) * n_embd_head,
ggml_element_size(tmpqkv_perm) * n_embd_head * n_head,
ggml_element_size(tmpqkv_perm) * n_embd_head * n_head * n_tokens * 2
);
cb(Vcur, "Vcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
model.layers[il].wo, model.layers[il].bo,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Q, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
}
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
residual = ggml_get_rows(ctx0, residual, inp_out_ids);
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * ffn_inp = ggml_add(ctx0, residual, cur);
cb(ffn_inp, "ffn_inp", il);
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// feed-forward network
{
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm,
model.layers[il].ffn_norm_b,
LLM_NORM, cb, il);
cb(cur, "ffn_norm", il);
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, model.layers[il].ffn_up_b,
NULL, NULL,
model.layers[il].ffn_down, model.layers[il].ffn_down_b,
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
NULL,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
LLM_FFN_RELU_SQR, LLM_FFN_SEQ, cb, il);
cb(cur, "ffn_out", il);
}
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "l_out", il);
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
inpL = cur;
}
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = inpL;
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm,
model.output_norm_b,
LLM_NORM, cb, -1);
cb(cur, "result_norm", -1);
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
ggml_build_forward_expand(gf, cur);
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
return gf;
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_cgraph * build_refact() {
sync : ggml (backend v2) (#3912) * sync : ggml (backend v2) (wip) * sync : migrate examples and llama.cpp to dynamic graphs (wip) * sync : update tests + fix max op params to 64 ggml-ci * sync : ggml-cuda ggml-ci * llama : fix save/load state context size ggml-ci * sync : try to fix build on tvOS * sync : pass custom graph sizes in training examples * sync : update graph copies to new ggml API * sync : update sync-ggml.sh with new files * scripts : fix header in sync script * train : fix context size calculations * llama : increase inference graph size up to 4096 nodes * train : allocate grads for backward graphs * train : allocate grads for gb_tmp
2023-11-13 12:16:23 +00:00
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
starcoder : add GPU offloading (#3827) * starcoder : do not GPU split 1D bias tensors * starcoder : offload layers to GPU ggml-ci
2023-10-28 09:06:08 +00:00
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_embd_head = hparams.n_embd_head_v;
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
starcoder : add GPU offloading (#3827) * starcoder : do not GPU split 1D bias tensors * starcoder : offload layers to GPU ggml-ci
2023-10-28 09:06:08 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
// positions of the tokens in the KV cache
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_pos = build_inp_KQ_pos();
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * inpSA = inpL;
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "attn_norm", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// self-attention
{
struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
cb(Qcur, "Qcur", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur);
cb(Kcur, "Kcur", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur);
cb(Vcur, "Vcur", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
cb(Kcur, "Kcur", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
cb(Qcur, "Qcur", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
model.layers[il].wo, NULL,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, KQ_pos, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
}
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
cb(ffn_inp, "ffn_inp", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// feed-forward network
{
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "ffn_norm", il);
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, NULL,
model.layers[il].ffn_gate, NULL,
model.layers[il].ffn_down, NULL,
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
NULL,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
cb(cur, "ffn_out", il);
}
starcoder : add GPU offloading (#3827) * starcoder : do not GPU split 1D bias tensors * starcoder : offload layers to GPU ggml-ci
2023-10-28 09:06:08 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "l_out", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// input for next layer
inpL = cur;
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = inpL;
starcoder : add GPU offloading (#3827) * starcoder : do not GPU split 1D bias tensors * starcoder : offload layers to GPU ggml-ci
2023-10-28 09:06:08 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm, NULL,
LLM_NORM_RMS, cb, -1);
cb(cur, "result_norm", -1);
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// lm_head
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
ggml_build_forward_expand(gf, cur);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
return gf;
}
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
struct ggml_cgraph * build_bert() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
const int64_t n_embd_head = hparams.n_embd_head_v;
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
const int64_t n_embd_gqa = hparams.n_embd_v_gqa();
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * inp_pos = build_inp_pos();
struct ggml_tensor * inp_mean = build_inp_mean();
struct ggml_tensor * inp_cls = build_inp_cls();
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
// construct input embeddings (token, type, position)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llama : support batched embeddings (#5466) * batched embedding: pool outputs by sequence id. updated embedding example * bring back non-causal attention * embd : minor improvements * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-13 12:06:58 +00:00
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
// token types are hardcoded to zero ("Sentence A")
struct ggml_tensor * type_row0 = ggml_view_1d(ctx0, model.type_embd, n_embd, 0);
inpL = ggml_add(ctx0, inpL, type_row0);
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
if (model.arch == LLM_ARCH_BERT) {
inpL = ggml_add(ctx0, ggml_get_rows(ctx0, model.pos_embd, inp_pos), inpL);
}
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
cb(inpL, "inp_embd", -1);
// embed layer norm
inpL = llm_build_norm(ctx0, inpL, hparams, model.tok_norm, model.tok_norm_b, LLM_NORM, cb, -1);
cb(inpL, "inp_norm", -1);
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask(false);
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
// iterate layers
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * cur = inpL;
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
struct ggml_tensor * Qcur;
struct ggml_tensor * Kcur;
struct ggml_tensor * Vcur;
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
// self-attention
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
if (model.arch == LLM_ARCH_BERT) {
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
Qcur = ggml_add(ctx0, ggml_mul_mat(ctx0, model.layers[il].wq, cur), model.layers[il].bq);
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
cb(Qcur, "Qcur", il);
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
Kcur = ggml_add(ctx0, ggml_mul_mat(ctx0, model.layers[il].wk, cur), model.layers[il].bk);
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
cb(Kcur, "Kcur", il);
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
Vcur = ggml_add(ctx0, ggml_mul_mat(ctx0, model.layers[il].wv, cur), model.layers[il].bv);
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
cb(Vcur, "Vcur", il);
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
} else {
// compute Q and K and RoPE them
cur = ggml_mul_mat(ctx0, model.layers[il].wqkv, cur);
cb(cur, "wqkv", il);
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
Qcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd, n_tokens, cur->nb[1], 0*sizeof(float)*(n_embd)));
Kcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd)));
Vcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd + n_embd_gqa)));
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
cb(Qcur, "Qcur", il);
cb(Kcur, "Kcur", il);
cb(Vcur, "Vcur", il);
Qcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Qcur, "Qcur", il);
Kcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Kcur, "Kcur", il);
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
}
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
struct ggml_tensor * q = ggml_permute(ctx0, Qcur, 0, 2, 1, 3);
struct ggml_tensor * k = ggml_cont(ctx0, ggml_permute(ctx0, Kcur, 0, 2, 1, 3));
struct ggml_tensor * kq = ggml_mul_mat(ctx0, k, q);
cb(kq, "kq", il);
kq = ggml_soft_max_ext(ctx0, kq, KQ_mask, nullptr, 1.0f/sqrtf(float(n_embd_head)), hparams.f_max_alibi_bias);
cb(kq, "kq_soft_max_ext", il);
struct ggml_tensor * v = ggml_cont(ctx0, ggml_transpose(ctx0, ggml_reshape_2d(ctx0, Vcur, n_embd_gqa, n_tokens)));
cb(v, "v", il);
struct ggml_tensor * kqv = ggml_mul_mat(ctx0, ggml_reshape_3d(ctx0, v, n_tokens, n_embd_head, n_head_kv), kq);
cb(kqv, "kqv", il);
struct ggml_tensor * kqv_merged = ggml_permute(ctx0, kqv, 0, 2, 1, 3);
cb(kqv_merged, "kqv_merged", il);
cur = ggml_cont_2d(ctx0, kqv_merged, n_embd_gqa, n_tokens);
cb(cur, "kqv_merged_cont", il);
ggml_build_forward_expand(gf, cur);
cur = ggml_mul_mat(ctx0, model.layers[il].wo, cur);
if (model.layers[il].bo) {
cb(cur, "kqv_wo", il);
}
if (model.layers[il].bo) {
cur = ggml_add(ctx0, cur, model.layers[il].bo);
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
}
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
cb(cur, "kqv_out", il);
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1 && pooling_type == LLAMA_POOLING_TYPE_NONE) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
}
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
// re-add the layer input
cur = ggml_add(ctx0, cur, inpL);
// attention layer norm
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
cur = llm_build_norm(ctx0, cur, hparams, model.layers[il].attn_out_norm, model.layers[il].attn_out_norm_b, LLM_NORM, cb, il);
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
struct ggml_tensor * ffn_inp = cur;
cb(ffn_inp, "ffn_inp", il);
// feed-forward network
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
if (model.arch == LLM_ARCH_BERT) {
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, model.layers[il].ffn_up_b,
NULL, NULL,
model.layers[il].ffn_down, model.layers[il].ffn_down_b,
NULL,
LLM_FFN_GELU, LLM_FFN_SEQ, cb, il);
} else {
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, NULL,
model.layers[il].ffn_gate, NULL,
model.layers[il].ffn_down, NULL,
NULL,
LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
}
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
cb(cur, "ffn_out", il);
// attentions bypass the intermediate layer
cur = ggml_add(ctx0, cur, ffn_inp);
// output layer norm
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
cur = llm_build_norm(ctx0, cur, hparams, model.layers[il].layer_out_norm, model.layers[il].layer_out_norm_b, LLM_NORM, cb, il);
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
// input for next layer
inpL = cur;
}
// final output
cur = inpL;
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
cb(cur, "result_embd", -1);
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
llama : support batched embeddings (#5466) * batched embedding: pool outputs by sequence id. updated embedding example * bring back non-causal attention * embd : minor improvements * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-13 12:06:58 +00:00
// pooling layer
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
switch (pooling_type) {
case LLAMA_POOLING_TYPE_NONE:
{
// nop
} break;
case LLAMA_POOLING_TYPE_MEAN:
{
cur = ggml_mul_mat(ctx0, ggml_cont(ctx0, ggml_transpose(ctx0, cur)), inp_mean);
cb(cur, "result_embd_pooled", -1);
} break;
case LLAMA_POOLING_TYPE_CLS:
{
cur = ggml_get_rows(ctx0, cur, inp_cls);
cb(cur, "result_embd_pooled", -1);
} break;
case LLAMA_POOLING_TYPE_UNSPECIFIED:
{
GGML_ASSERT(false && "Invalid pooling type");
} break;
llama : support batched embeddings (#5466) * batched embedding: pool outputs by sequence id. updated embedding example * bring back non-causal attention * embd : minor improvements * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-13 12:06:58 +00:00
}
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
ggml_build_forward_expand(gf, cur);
return gf;
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_cgraph * build_bloom() {
sync : ggml (backend v2) (#3912) * sync : ggml (backend v2) (wip) * sync : migrate examples and llama.cpp to dynamic graphs (wip) * sync : update tests + fix max op params to 64 ggml-ci * sync : ggml-cuda ggml-ci * llama : fix save/load state context size ggml-ci * sync : try to fix build on tvOS * sync : pass custom graph sizes in training examples * sync : update graph copies to new ggml API * sync : update sync-ggml.sh with new files * scripts : fix header in sync script * train : fix context size calculations * llama : increase inference graph size up to 4096 nodes * train : allocate grads for backward graphs * train : allocate grads for gb_tmp
2023-11-13 12:16:23 +00:00
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_embd_head = hparams.n_embd_head_v;
const int64_t n_embd_gqa = hparams.n_embd_v_gqa();
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
// positions of the tokens in the KV cache
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_pos = build_inp_KQ_pos();
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
inpL = llm_build_norm(ctx0, inpL, hparams,
model.tok_norm,
model.tok_norm_b,
LLM_NORM, cb, -1);
cb(inpL, "inp_norm", -1);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
for (int il = 0; il < n_layer; ++il) {
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm,
model.layers[il].attn_norm_b,
LLM_NORM, cb, il);
cb(cur, "attn_norm", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// self-attention
{
cur = ggml_mul_mat(ctx0, model.layers[il].wqkv, cur);
cb(cur, "wqkv", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = ggml_add(ctx0, cur, model.layers[il].bqkv);
cb(cur, "bqkv", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * Qcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd, n_tokens, cur->nb[1], 0*sizeof(float)*(n_embd)));
struct ggml_tensor * Kcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd)));
struct ggml_tensor * Vcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd + n_embd_gqa)));
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cb(Qcur, "Qcur", il);
cb(Kcur, "Kcur", il);
cb(Vcur, "Vcur", il);
starcoder : add GPU offloading (#3827) * starcoder : do not GPU split 1D bias tensors * starcoder : offload layers to GPU ggml-ci
2023-10-28 09:06:08 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
model.layers[il].wo, model.layers[il].bo,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, KQ_pos, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
}
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// Add the input
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL);
cb(ffn_inp, "ffn_inp", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// FF
{
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm,
model.layers[il].ffn_norm_b,
LLM_NORM, cb, il);
cb(cur, "ffn_norm", il);
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, model.layers[il].ffn_up_b,
NULL, NULL,
model.layers[il].ffn_down, model.layers[il].ffn_down_b,
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
NULL,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
LLM_FFN_GELU, LLM_FFN_SEQ, cb, il);
cb(cur, "ffn_out", il);
}
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
inpL = ggml_add(ctx0, cur, ffn_inp);
cb(inpL, "l_out", il);
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = llm_build_norm(ctx0, inpL, hparams,
model.output_norm,
model.output_norm_b,
LLM_NORM, cb, -1);
cb(cur, "result_norm", -1);
starcoder : add GPU offloading (#3827) * starcoder : do not GPU split 1D bias tensors * starcoder : offload layers to GPU ggml-ci
2023-10-28 09:06:08 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
starcoder : add GPU offloading (#3827) * starcoder : do not GPU split 1D bias tensors * starcoder : offload layers to GPU ggml-ci
2023-10-28 09:06:08 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
ggml_build_forward_expand(gf, cur);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
return gf;
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_cgraph * build_mpt() {
sync : ggml (backend v2) (#3912) * sync : ggml (backend v2) (wip) * sync : migrate examples and llama.cpp to dynamic graphs (wip) * sync : update tests + fix max op params to 64 ggml-ci * sync : ggml-cuda ggml-ci * llama : fix save/load state context size ggml-ci * sync : try to fix build on tvOS * sync : pass custom graph sizes in training examples * sync : update graph copies to new ggml API * sync : update sync-ggml.sh with new files * scripts : fix header in sync script * train : fix context size calculations * llama : increase inference graph size up to 4096 nodes * train : allocate grads for backward graphs * train : allocate grads for gb_tmp
2023-11-13 12:16:23 +00:00
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_embd_head = hparams.n_embd_head_v;
const int64_t n_embd_gqa = hparams.n_embd_v_gqa();
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * cur;
llama : add SEA-LION support (#6448) * initial commit for sealion support * add sealion support * minor fix * q/k ln and pos_embd only if required * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * minor : clear whitespaces --------- Co-authored-by: bryan <bryansiow@aisingapore.org> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 18:05:10 +00:00
struct ggml_tensor * pos;
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * inpL;
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
// positions of the tokens in the KV cache
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_pos = build_inp_KQ_pos();
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
llama : add SEA-LION support (#6448) * initial commit for sealion support * add sealion support * minor fix * q/k ln and pos_embd only if required * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * minor : clear whitespaces --------- Co-authored-by: bryan <bryansiow@aisingapore.org> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 18:05:10 +00:00
if (model.pos_embd) {
// inp_pos - contains the positions
struct ggml_tensor * inp_pos = build_inp_pos();
pos = ggml_get_rows(ctx0, model.pos_embd, inp_pos);
cb(pos, "pos_embd", -1);
inpL = ggml_add(ctx0, inpL, pos);
cb(inpL, "inpL", -1);
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * attn_norm;
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
attn_norm = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm,
mpt : add optional bias tensors (#5638) Update for MPT with optional bias parameters: to work with PhoGPT and SEA-LION models that were pre-trained with 'bias'.
2024-02-22 08:15:13 +00:00
model.layers[il].attn_norm_b,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
LLM_NORM, cb, il);
cb(attn_norm, "attn_norm", il);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// self-attention
{
cur = attn_norm;
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = ggml_mul_mat(ctx0, model.layers[il].wqkv, cur);
cb(cur, "wqkv", il);
minor : fix trailing whitespace (#5638)
2024-02-22 11:54:03 +00:00
mpt : add optional bias tensors (#5638) Update for MPT with optional bias parameters: to work with PhoGPT and SEA-LION models that were pre-trained with 'bias'.
2024-02-22 08:15:13 +00:00
if (model.layers[il].bqkv){
cur = ggml_add(ctx0, cur, model.layers[il].bqkv);
cb(cur, "bqkv", il);
}
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
if (hparams.f_clamp_kqv > 0.0f) {
cur = ggml_clamp(ctx0, cur, -hparams.f_clamp_kqv, hparams.f_clamp_kqv);
cb(cur, "wqkv_clamped", il);
}
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
struct ggml_tensor * Qcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd, n_tokens, cur->nb[1], 0*sizeof(float)*(n_embd)));
struct ggml_tensor * Kcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd)));
struct ggml_tensor * Vcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd + n_embd_gqa)));
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cb(Qcur, "Qcur", il);
cb(Kcur, "Kcur", il);
cb(Vcur, "Vcur", il);
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
llama : add SEA-LION support (#6448) * initial commit for sealion support * add sealion support * minor fix * q/k ln and pos_embd only if required * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * minor : clear whitespaces --------- Co-authored-by: bryan <bryansiow@aisingapore.org> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 18:05:10 +00:00
// Q/K Layernorm
if (model.layers[il].attn_q_norm) {
Qcur = llm_build_norm(ctx0, Qcur, hparams,
model.layers[il].attn_q_norm,
model.layers[il].attn_q_norm_b,
LLM_NORM, cb, il);
cb(Qcur, "Qcur", il);
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
llama : add SEA-LION support (#6448) * initial commit for sealion support * add sealion support * minor fix * q/k ln and pos_embd only if required * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * minor : clear whitespaces --------- Co-authored-by: bryan <bryansiow@aisingapore.org> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 18:05:10 +00:00
Kcur = llm_build_norm(ctx0, Kcur, hparams,
model.layers[il].attn_k_norm,
model.layers[il].attn_k_norm_b,
LLM_NORM, cb, il);
cb(Kcur, "Kcur", il);
Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
model.layers[il].wo, model.layers[il].bo,
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llama : add SEA-LION support (#6448) * initial commit for sealion support * add sealion support * minor fix * q/k ln and pos_embd only if required * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * minor : clear whitespaces --------- Co-authored-by: bryan <bryansiow@aisingapore.org> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 18:05:10 +00:00
} else {
Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llama : add SEA-LION support (#6448) * initial commit for sealion support * add sealion support * minor fix * q/k ln and pos_embd only if required * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * minor : clear whitespaces --------- Co-authored-by: bryan <bryansiow@aisingapore.org> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 18:05:10 +00:00
model.layers[il].wo, model.layers[il].bo,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, KQ_pos, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llama : add SEA-LION support (#6448) * initial commit for sealion support * add sealion support * minor fix * q/k ln and pos_embd only if required * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * minor : clear whitespaces --------- Co-authored-by: bryan <bryansiow@aisingapore.org> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 18:05:10 +00:00
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
}
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// Add the input
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL);
cb(ffn_inp, "ffn_inp", il);
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// feed forward
{
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm,
mpt : add optional bias tensors (#5638) Update for MPT with optional bias parameters: to work with PhoGPT and SEA-LION models that were pre-trained with 'bias'.
2024-02-22 08:15:13 +00:00
model.layers[il].ffn_norm_b,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
LLM_NORM, cb, il);
cb(cur, "ffn_norm", il);
cur = llm_build_ffn(ctx0, cur,
mpt : add optional bias tensors (#5638) Update for MPT with optional bias parameters: to work with PhoGPT and SEA-LION models that were pre-trained with 'bias'.
2024-02-22 08:15:13 +00:00
model.layers[il].ffn_up, model.layers[il].ffn_up_b,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
NULL, NULL,
mpt : add optional bias tensors (#5638) Update for MPT with optional bias parameters: to work with PhoGPT and SEA-LION models that were pre-trained with 'bias'.
2024-02-22 08:15:13 +00:00
model.layers[il].ffn_down, model.layers[il].ffn_down_b,
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
model.layers[il].ffn_act,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
LLM_FFN_GELU, LLM_FFN_SEQ, cb, il);
cb(cur, "ffn_out", il);
}
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "l_out", il);
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
// input for next layer
inpL = cur;
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
}
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = inpL;
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm,
mpt : add optional bias tensors (#5638) Update for MPT with optional bias parameters: to work with PhoGPT and SEA-LION models that were pre-trained with 'bias'.
2024-02-22 08:15:13 +00:00
model.output_norm_b,
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
LLM_NORM, cb, -1);
cb(cur, "result_norm", -1);
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
ggml_build_forward_expand(gf, cur);
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
return gf;
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
}
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
struct ggml_cgraph * build_stablelm() {
struct ggml_cgraph * gf = ggml_new_graph(ctx0);
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_embd_head = hparams.n_embd_head_v;
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
// inp_pos - contains the positions
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * inp_pos = build_inp_pos();
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
for (int il = 0; il < n_layer; ++il) {
llama : add StableLM2 12B (#6635) * StableLM2 12B support for huggingface -> GGUF * StableLM12 tensormapping and constants * StableLM-2-12b model support * fix * Added 12B support * Removed autoformatting; resolved bug where model_arch was not selecting StableLM2 * Formatting * Do QK norm stacking in model conversion step * Converge StableLM and StableLM2 code to simplify graph construction * Fix accidental removal * Removed warnings * Revert formatter * Move QK norm stack to private function so it's easier to read * refactor stablelm graph builder to support 1.6, 3b and 12b more efficiently * Proper check for None type for new_name to avoid crash; formatting; revert change to base class `write_tensors()` * Format * Formatting * format Co-authored-by: compilade <git@compilade.net> * Fix incorrect check for K norm * space after commas; Keep indentation multiple of 4 spaces * Flake8 format * Removed unnecessary conditional branches * Removed unused comment * Fixed incorrect tensor passing * Format --------- Co-authored-by: compilade <git@compilade.net>
2024-04-16 15:48:35 +00:00
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
// norm
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm,
model.layers[il].attn_norm_b,
LLM_NORM, cb, il);
cb(cur, "attn_norm", il);
llama : add StableLM2 12B (#6635) * StableLM2 12B support for huggingface -> GGUF * StableLM12 tensormapping and constants * StableLM-2-12b model support * fix * Added 12B support * Removed autoformatting; resolved bug where model_arch was not selecting StableLM2 * Formatting * Do QK norm stacking in model conversion step * Converge StableLM and StableLM2 code to simplify graph construction * Fix accidental removal * Removed warnings * Revert formatter * Move QK norm stack to private function so it's easier to read * refactor stablelm graph builder to support 1.6, 3b and 12b more efficiently * Proper check for None type for new_name to avoid crash; formatting; revert change to base class `write_tensors()` * Format * Formatting * format Co-authored-by: compilade <git@compilade.net> * Fix incorrect check for K norm * space after commas; Keep indentation multiple of 4 spaces * Flake8 format * Removed unnecessary conditional branches * Removed unused comment * Fixed incorrect tensor passing * Format --------- Co-authored-by: compilade <git@compilade.net>
2024-04-16 15:48:35 +00:00
struct ggml_tensor * inpSA = cur;
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
// self-attention
{
// compute Q and K and RoPE them
stablelm : simplify + speedup generation (#4153)
2023-11-21 15:22:30 +00:00
struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
cb(Qcur, "Qcur", il);
llama : support StableLM 2 1.6B (#5052) * llama : support StableLM 2 1.6B * convert : fix Qwen's set_vocab wrongly naming all special tokens [PAD{id}] * convert : refactor Qwen's set_vocab to use it for StableLM 2 too * nix : add tiktoken to llama-python-extra * convert : use presence of tokenizer.json to determine StableLM tokenizer loader It's a less arbitrary heuristic than the vocab size.
2024-01-22 11:21:52 +00:00
if (model.layers[il].bq) {
Qcur = ggml_add(ctx0, Qcur, model.layers[il].bq);
cb(Qcur, "Qcur", il);
}
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
stablelm : simplify + speedup generation (#4153)
2023-11-21 15:22:30 +00:00
struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur);
cb(Kcur, "Kcur", il);
llama : support StableLM 2 1.6B (#5052) * llama : support StableLM 2 1.6B * convert : fix Qwen's set_vocab wrongly naming all special tokens [PAD{id}] * convert : refactor Qwen's set_vocab to use it for StableLM 2 too * nix : add tiktoken to llama-python-extra * convert : use presence of tokenizer.json to determine StableLM tokenizer loader It's a less arbitrary heuristic than the vocab size.
2024-01-22 11:21:52 +00:00
if (model.layers[il].bk) {
Kcur = ggml_add(ctx0, Kcur, model.layers[il].bk);
cb(Kcur, "Kcur", il);
}
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur);
cb(Vcur, "Vcur", il);
llama : support StableLM 2 1.6B (#5052) * llama : support StableLM 2 1.6B * convert : fix Qwen's set_vocab wrongly naming all special tokens [PAD{id}] * convert : refactor Qwen's set_vocab to use it for StableLM 2 too * nix : add tiktoken to llama-python-extra * convert : use presence of tokenizer.json to determine StableLM tokenizer loader It's a less arbitrary heuristic than the vocab size.
2024-01-22 11:21:52 +00:00
if (model.layers[il].bv) {
Vcur = ggml_add(ctx0, Vcur, model.layers[il].bv);
cb(Vcur, "Vcur", il);
}
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
llama : add StableLM2 12B (#6635) * StableLM2 12B support for huggingface -> GGUF * StableLM12 tensormapping and constants * StableLM-2-12b model support * fix * Added 12B support * Removed autoformatting; resolved bug where model_arch was not selecting StableLM2 * Formatting * Do QK norm stacking in model conversion step * Converge StableLM and StableLM2 code to simplify graph construction * Fix accidental removal * Removed warnings * Revert formatter * Move QK norm stack to private function so it's easier to read * refactor stablelm graph builder to support 1.6, 3b and 12b more efficiently * Proper check for None type for new_name to avoid crash; formatting; revert change to base class `write_tensors()` * Format * Formatting * format Co-authored-by: compilade <git@compilade.net> * Fix incorrect check for K norm * space after commas; Keep indentation multiple of 4 spaces * Flake8 format * Removed unnecessary conditional branches * Removed unused comment * Fixed incorrect tensor passing * Format --------- Co-authored-by: compilade <git@compilade.net>
2024-04-16 15:48:35 +00:00
Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
cb(Qcur, "Qcur", il);
Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
cb(Kcur, "Kcur", il);
if (model.layers[il].attn_q_norm) {
Qcur = llm_build_norm(ctx0, Qcur, hparams,
model.layers[il].attn_q_norm,
NULL,
LLM_NORM, cb, il);
cb(Qcur, "Qcur", il);
}
if (model.layers[il].attn_k_norm) {
Kcur = llm_build_norm(ctx0, Kcur, hparams,
model.layers[il].attn_k_norm,
NULL,
LLM_NORM, cb, il);
cb(Kcur, "Kcur", il);
}
stablelm : simplify + speedup generation (#4153)
2023-11-21 15:22:30 +00:00
Qcur = ggml_rope_custom(
llama : add StableLM2 12B (#6635) * StableLM2 12B support for huggingface -> GGUF * StableLM12 tensormapping and constants * StableLM-2-12b model support * fix * Added 12B support * Removed autoformatting; resolved bug where model_arch was not selecting StableLM2 * Formatting * Do QK norm stacking in model conversion step * Converge StableLM and StableLM2 code to simplify graph construction * Fix accidental removal * Removed warnings * Revert formatter * Move QK norm stack to private function so it's easier to read * refactor stablelm graph builder to support 1.6, 3b and 12b more efficiently * Proper check for None type for new_name to avoid crash; formatting; revert change to base class `write_tensors()` * Format * Formatting * format Co-authored-by: compilade <git@compilade.net> * Fix incorrect check for K norm * space after commas; Keep indentation multiple of 4 spaces * Flake8 format * Removed unnecessary conditional branches * Removed unused comment * Fixed incorrect tensor passing * Format --------- Co-authored-by: compilade <git@compilade.net>
2024-04-16 15:48:35 +00:00
ctx0, Qcur, inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
stablelm : simplify + speedup generation (#4153)
2023-11-21 15:22:30 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
);
cb(Qcur, "Qcur", il);
stablelm : simplify + speedup generation (#4153)
2023-11-21 15:22:30 +00:00
Kcur = ggml_rope_custom(
llama : add StableLM2 12B (#6635) * StableLM2 12B support for huggingface -> GGUF * StableLM12 tensormapping and constants * StableLM-2-12b model support * fix * Added 12B support * Removed autoformatting; resolved bug where model_arch was not selecting StableLM2 * Formatting * Do QK norm stacking in model conversion step * Converge StableLM and StableLM2 code to simplify graph construction * Fix accidental removal * Removed warnings * Revert formatter * Move QK norm stack to private function so it's easier to read * refactor stablelm graph builder to support 1.6, 3b and 12b more efficiently * Proper check for None type for new_name to avoid crash; formatting; revert change to base class `write_tensors()` * Format * Formatting * format Co-authored-by: compilade <git@compilade.net> * Fix incorrect check for K norm * space after commas; Keep indentation multiple of 4 spaces * Flake8 format * Removed unnecessary conditional branches * Removed unused comment * Fixed incorrect tensor passing * Format --------- Co-authored-by: compilade <git@compilade.net>
2024-04-16 15:48:35 +00:00
ctx0, Kcur, inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
stablelm : simplify + speedup generation (#4153)
2023-11-21 15:22:30 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
);
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
model.layers[il].wo, NULL,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
llama : add StableLM2 12B (#6635) * StableLM2 12B support for huggingface -> GGUF * StableLM12 tensormapping and constants * StableLM-2-12b model support * fix * Added 12B support * Removed autoformatting; resolved bug where model_arch was not selecting StableLM2 * Formatting * Do QK norm stacking in model conversion step * Converge StableLM and StableLM2 code to simplify graph construction * Fix accidental removal * Removed warnings * Revert formatter * Move QK norm stack to private function so it's easier to read * refactor stablelm graph builder to support 1.6, 3b and 12b more efficiently * Proper check for None type for new_name to avoid crash; formatting; revert change to base class `write_tensors()` * Format * Formatting * format Co-authored-by: compilade <git@compilade.net> * Fix incorrect check for K norm * space after commas; Keep indentation multiple of 4 spaces * Flake8 format * Removed unnecessary conditional branches * Removed unused comment * Fixed incorrect tensor passing * Format --------- Co-authored-by: compilade <git@compilade.net>
2024-04-16 15:48:35 +00:00
inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
}
llama : add StableLM2 12B (#6635) * StableLM2 12B support for huggingface -> GGUF * StableLM12 tensormapping and constants * StableLM-2-12b model support * fix * Added 12B support * Removed autoformatting; resolved bug where model_arch was not selecting StableLM2 * Formatting * Do QK norm stacking in model conversion step * Converge StableLM and StableLM2 code to simplify graph construction * Fix accidental removal * Removed warnings * Revert formatter * Move QK norm stack to private function so it's easier to read * refactor stablelm graph builder to support 1.6, 3b and 12b more efficiently * Proper check for None type for new_name to avoid crash; formatting; revert change to base class `write_tensors()` * Format * Formatting * format Co-authored-by: compilade <git@compilade.net> * Fix incorrect check for K norm * space after commas; Keep indentation multiple of 4 spaces * Flake8 format * Removed unnecessary conditional branches * Removed unused comment * Fixed incorrect tensor passing * Format --------- Co-authored-by: compilade <git@compilade.net>
2024-04-16 15:48:35 +00:00
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL);
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
cb(ffn_inp, "ffn_inp", il);
// feed-forward network
{
llama : add StableLM2 12B (#6635) * StableLM2 12B support for huggingface -> GGUF * StableLM12 tensormapping and constants * StableLM-2-12b model support * fix * Added 12B support * Removed autoformatting; resolved bug where model_arch was not selecting StableLM2 * Formatting * Do QK norm stacking in model conversion step * Converge StableLM and StableLM2 code to simplify graph construction * Fix accidental removal * Removed warnings * Revert formatter * Move QK norm stack to private function so it's easier to read * refactor stablelm graph builder to support 1.6, 3b and 12b more efficiently * Proper check for None type for new_name to avoid crash; formatting; revert change to base class `write_tensors()` * Format * Formatting * format Co-authored-by: compilade <git@compilade.net> * Fix incorrect check for K norm * space after commas; Keep indentation multiple of 4 spaces * Flake8 format * Removed unnecessary conditional branches * Removed unused comment * Fixed incorrect tensor passing * Format --------- Co-authored-by: compilade <git@compilade.net>
2024-04-16 15:48:35 +00:00
if (model.layers[il].ffn_norm) {
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm,
model.layers[il].ffn_norm_b,
LLM_NORM, cb, il);
cb(cur, "ffn_norm", il);
} else {
// parallel residual
cur = inpSA;
}
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, NULL,
model.layers[il].ffn_gate, NULL,
model.layers[il].ffn_down, NULL,
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
NULL,
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
cb(cur, "ffn_out", il);
}
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "l_out", il);
// input for next layer
inpL = cur;
}
cur = inpL;
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm,
model.output_norm_b,
LLM_NORM, cb, -1);
cb(cur, "result_norm", -1);
// lm_head
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
return gf;
}
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
struct ggml_cgraph * build_qwen() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_embd_head = hparams.n_embd_head_v;
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
// inp_pos - contains the positions
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * inp_pos = build_inp_pos();
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * inpSA = inpL;
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "attn_norm", il);
// self-attention
{
cur = ggml_mul_mat(ctx0, model.layers[il].wqkv, cur);
cb(cur, "wqkv", il);
cur = ggml_add(ctx0, cur, model.layers[il].bqkv);
cb(cur, "bqkv", il);
struct ggml_tensor * Qcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd, n_tokens, cur->nb[1], 0*sizeof(float)*(n_embd)));
struct ggml_tensor * Kcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd)));
struct ggml_tensor * Vcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd, n_tokens, cur->nb[1], 2*sizeof(float)*(n_embd)));
cb(Qcur, "Qcur", il);
cb(Kcur, "Kcur", il);
cb(Vcur, "Vcur", il);
Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
// using mode = 2 for neox mode
Qcur = ggml_rope_custom(
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ctx0, Qcur, inp_pos, n_rot, rope_type, 0, n_orig_ctx,
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
freq_base, freq_scale, ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Qcur, "Qcur", il);
Kcur = ggml_rope_custom(
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ctx0, Kcur, inp_pos, n_rot, rope_type, 0, n_orig_ctx,
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
freq_base, freq_scale, ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
model.layers[il].wo, NULL,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
}
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
cb(ffn_inp, "ffn_inp", il);
// feed-forward forward
{
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "ffn_norm", il);
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, NULL,
model.layers[il].ffn_gate, NULL,
model.layers[il].ffn_down, NULL,
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
NULL,
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
cb(cur, "ffn_out", il);
}
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "l_out", il);
// input for next layer
inpL = cur;
}
cur = inpL;
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm, NULL,
LLM_NORM_RMS, cb, -1);
cb(cur, "result_norm", -1);
// lm_head
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
return gf;
}
llama : support upcoming Qwen2 (#5037)
2024-01-19 11:53:13 +00:00
struct ggml_cgraph * build_qwen2() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
const int64_t n_embd_head = hparams.n_embd_head_v;
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
GGML_ASSERT(n_embd_head == hparams.n_rot);
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llama : support upcoming Qwen2 (#5037)
2024-01-19 11:53:13 +00:00
// inp_pos - contains the positions
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * inp_pos = build_inp_pos();
llama : support upcoming Qwen2 (#5037)
2024-01-19 11:53:13 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llama : support upcoming Qwen2 (#5037)
2024-01-19 11:53:13 +00:00
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * inpSA = inpL;
// norm
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "attn_norm", il);
// self-attention
{
// compute Q and K and RoPE them
struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
cb(Qcur, "Qcur", il);
Qcur = ggml_add(ctx0, Qcur, model.layers[il].bq);
cb(Qcur, "Qcur", il);
struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur);
cb(Kcur, "Kcur", il);
Kcur = ggml_add(ctx0, Kcur, model.layers[il].bk);
cb(Kcur, "Kcur", il);
struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur);
cb(Vcur, "Vcur", il);
Vcur = ggml_add(ctx0, Vcur, model.layers[il].bv);
cb(Vcur, "Vcur", il);
Qcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : support upcoming Qwen2 (#5037)
2024-01-19 11:53:13 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Qcur, "Qcur", il);
Kcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : support upcoming Qwen2 (#5037)
2024-01-19 11:53:13 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llama : support upcoming Qwen2 (#5037)
2024-01-19 11:53:13 +00:00
model.layers[il].wo, model.layers[il].bo,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llama : support upcoming Qwen2 (#5037)
2024-01-19 11:53:13 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
}
llama : support upcoming Qwen2 (#5037)
2024-01-19 11:53:13 +00:00
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
cb(ffn_inp, "ffn_inp", il);
// feed-forward network
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "ffn_norm", il);
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, NULL,
model.layers[il].ffn_gate, NULL,
model.layers[il].ffn_down, NULL,
NULL,
LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
cb(cur, "ffn_out", il);
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "l_out", il);
// input for next layer
inpL = cur;
}
cur = inpL;
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm, NULL,
LLM_NORM_RMS, cb, -1);
cb(cur, "result_norm", -1);
// lm_head
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
return gf;
}
llama : add qwen2moe (#6074) * support qwen2moe * fix-review * metal : support unary ops for nelements % 4 != 0 * metal : require contiguousness for float4 unary kernels * metal : require contiguousness for float4 unary kernels (cont) * fix-review * names : for brevity "SHARED_EXP" -> "SHEXP" * llama : reuse build_moe_ffn() * llama : add model type name --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-16 15:40:48 +00:00
struct ggml_cgraph * build_qwen2moe() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
// mutable variable, needed during the last layer of the computation to skip unused tokens
int32_t n_tokens = this->n_tokens;
const int64_t n_embd_head = hparams.n_embd_head_v;
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
GGML_ASSERT(n_embd_head == hparams.n_rot);
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
// inp_pos - contains the positions
struct ggml_tensor * inp_pos = build_inp_pos();
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * inpSA = inpL;
// norm
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "attn_norm", il);
// self_attention
{
// compute Q and K and RoPE them
struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
cb(Qcur, "Qcur", il);
Qcur = ggml_add(ctx0, Qcur, model.layers[il].bq);
cb(Qcur, "Qcur", il);
struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur);
cb(Kcur, "Kcur", il);
Kcur = ggml_add(ctx0, Kcur, model.layers[il].bk);
cb(Kcur, "Kcur", il);
struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur);
cb(Vcur, "Vcur", il);
Vcur = ggml_add(ctx0, Vcur, model.layers[il].bv);
cb(Vcur, "Vcur", il);
Qcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens), inp_pos,
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Qcur, "Qcur", il);
Kcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens), inp_pos,
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llama : add qwen2moe (#6074) * support qwen2moe * fix-review * metal : support unary ops for nelements % 4 != 0 * metal : require contiguousness for float4 unary kernels * metal : require contiguousness for float4 unary kernels (cont) * fix-review * names : for brevity "SHARED_EXP" -> "SHEXP" * llama : reuse build_moe_ffn() * llama : add model type name --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-16 15:40:48 +00:00
model.layers[il].wo, model.layers[il].bo,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llama : add qwen2moe (#6074) * support qwen2moe * fix-review * metal : support unary ops for nelements % 4 != 0 * metal : require contiguousness for float4 unary kernels * metal : require contiguousness for float4 unary kernels (cont) * fix-review * names : for brevity "SHARED_EXP" -> "SHEXP" * llama : reuse build_moe_ffn() * llama : add model type name --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-16 15:40:48 +00:00
}
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
n_tokens = n_outputs;
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
}
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
cb(ffn_inp, "ffn_inp", il);
// MoE branch
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "ffn_norm", il);
ggml : group all experts in a single ggml_mul_mat_id (#6505) * ggml : group all experts in a single ggml_mul_mat_id cuda : improve mmid row copy * cuda : fix bin bcast with non-cont src0 * test-backend-ops : only run all mul mat tests for base types * llama : disable moe offloading with SYCL --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-18 13:18:48 +00:00
ggml_tensor * moe_out =
llm_build_moe_ffn(ctx0, cur,
model.layers[il].ffn_gate_inp,
model.layers[il].ffn_up_exps,
model.layers[il].ffn_gate_exps,
model.layers[il].ffn_down_exps,
n_expert, n_expert_used,
LLM_FFN_SILU, false,
cb, il);
cb(cur, "ffn_moe_out", il);
llama : add qwen2moe (#6074) * support qwen2moe * fix-review * metal : support unary ops for nelements % 4 != 0 * metal : require contiguousness for float4 unary kernels * metal : require contiguousness for float4 unary kernels (cont) * fix-review * names : for brevity "SHARED_EXP" -> "SHEXP" * llama : reuse build_moe_ffn() * llama : add model type name --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-16 15:40:48 +00:00
// FFN shared expert
{
ggml_tensor * cur_gate_inp = ggml_mul_mat(ctx0, model.layers[il].ffn_gate_inp_shexp, cur);
cb(cur_gate_inp, "ffn_shexp_gate_inp", il);
// sigmoid
ggml_tensor * cur_gate = ggml_div(ctx0, ggml_silu(ctx0, cur_gate_inp), cur_gate_inp);
cb(cur_gate, "ffn_shexp_gate", il);
ggml_tensor * cur_ffn = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up_shexp, NULL,
model.layers[il].ffn_gate_shexp, NULL,
model.layers[il].ffn_down_shexp, NULL,
NULL,
LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
cb(cur_ffn, "ffn_shexp", il);
ggml_tensor * ffn_shexp_out = ggml_mul(ctx0, cur_ffn, cur_gate);
cb(ffn_shexp_out, "ffn_shexp_out", il);
moe_out = ggml_add(ctx0, moe_out, ffn_shexp_out);
cb(moe_out, "ffn_out", il);
cur = moe_out;
}
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "l_out", il);
// input for next layer
inpL = cur;
}
cur = inpL;
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm, NULL,
LLM_NORM_RMS, cb, -1);
cb(cur, "result_norm", -1);
// lm_head
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
return gf;
}
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
struct ggml_cgraph * build_phi2() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_embd_head = hparams.n_embd_head_v;
const int64_t n_embd_gqa = hparams.n_embd_v_gqa();
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
struct ggml_tensor * cur;
struct ggml_tensor * attn_norm_output;
struct ggml_tensor * ffn_output;
struct ggml_tensor * inpL;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
// inp_pos - contains the positions
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * inp_pos = build_inp_pos();
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
for (int il = 0; il < n_layer; ++il) {
attn_norm_output = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm,
model.layers[il].attn_norm_b,
LLM_NORM, cb, il);
cb(attn_norm_output, "attn_norm", il);
// self-attention
{
convert : update phi-2 to latest HF repo (#4903) * convert : update phi-2 to latest HF repo ggml-ci * py : try to fix flake stuff
2024-01-13 11:44:37 +00:00
struct ggml_tensor * Qcur = nullptr;
struct ggml_tensor * Kcur = nullptr;
struct ggml_tensor * Vcur = nullptr;
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
convert : update phi-2 to latest HF repo (#4903) * convert : update phi-2 to latest HF repo ggml-ci * py : try to fix flake stuff
2024-01-13 11:44:37 +00:00
if (model.layers[il].wqkv) {
cur = ggml_mul_mat(ctx0, model.layers[il].wqkv, attn_norm_output);
cb(cur, "wqkv", il);
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
convert : update phi-2 to latest HF repo (#4903) * convert : update phi-2 to latest HF repo ggml-ci * py : try to fix flake stuff
2024-01-13 11:44:37 +00:00
cur = ggml_add(ctx0, cur, model.layers[il].bqkv);
cb(cur, "bqkv", il);
Qcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd, n_tokens, cur->nb[1], 0*sizeof(float)*(n_embd)));
Kcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd)));
Vcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd + n_embd_gqa)));
} else {
Qcur = ggml_add(ctx0, ggml_mul_mat(ctx0, model.layers[il].wq, attn_norm_output), model.layers[il].bq);
Kcur = ggml_add(ctx0, ggml_mul_mat(ctx0, model.layers[il].wk, attn_norm_output), model.layers[il].bk);
Vcur = ggml_add(ctx0, ggml_mul_mat(ctx0, model.layers[il].wv, attn_norm_output), model.layers[il].bv);
}
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
cb(Qcur, "Qcur", il);
cb(Kcur, "Kcur", il);
cb(Vcur, "Vcur", il);
Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
Qcur = ggml_rope_custom(
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ctx0, Qcur, inp_pos, n_rot, rope_type, 0, n_orig_ctx,
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
freq_base, freq_scale, ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Qcur, "Qcur", il);
ggml : change ggml_scale to take a float instead of tensor (#4573) * ggml : change ggml_scale to take a float instead of tensor * ggml : fix CPU implementation * tests : fix test-grad0 ggml-ci
2023-12-21 21:20:49 +00:00
// with phi2, we scale the Q to avoid precision issues
// ref: https://github.com/ml-explore/mlx-examples/blob/08e862336ade809bc37d1035f94b359e7d1a5152/phi2/phi2.py#L64-L66
Qcur = ggml_scale(ctx0, Qcur, 1.0f/sqrtf(float(n_embd_head)));
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
cb(Qcur, "Qcur", il);
Kcur = ggml_rope_custom(
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ctx0, Kcur, inp_pos, n_rot, rope_type, 0, n_orig_ctx,
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
freq_base, freq_scale, ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
model.layers[il].wo, model.layers[il].bo,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f, cb, il);
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
attn_norm_output = ggml_get_rows(ctx0, attn_norm_output, inp_out_ids);
}
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
// FF
{
ffn_output = llm_build_ffn(ctx0, attn_norm_output,
model.layers[il].ffn_up, model.layers[il].ffn_up_b,
NULL, NULL,
model.layers[il].ffn_down, model.layers[il].ffn_down_b,
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
NULL,
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
LLM_FFN_GELU, LLM_FFN_SEQ, cb, il);
cb(ffn_output, "ffn_out", il);
}
cur = ggml_add(ctx0, cur, ffn_output);
cb(cur, "l_out", il);
cur = ggml_add(ctx0, cur, inpL);
cb(cur, "l_out", il);
inpL = cur;
}
cur = llm_build_norm(ctx0, inpL, hparams,
model.output_norm,
model.output_norm_b,
LLM_NORM, cb, -1);
cb(cur, "result_norm", -1);
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output_no_bias", -1);
cur = ggml_add(ctx0, cur, model.output_b);
cb(cur, "result_output", -1);
llama : add phi3 support (#6852) * add explicit phi3 support * add explicit phi3 support * remove unused code * convert : add BOS token * llama : match EOT token <|end|> * llama : minor / style * llama : tabs -> spaces * convert : fix lint checks --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-24 07:00:37 +00:00
ggml_build_forward_expand(gf, cur);
return gf;
}
struct ggml_cgraph * build_phi3() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
const int64_t n_embd_head = hparams.n_embd_head_v;
const int64_t n_embd_gqa = hparams.n_embd_v_gqa();
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
// inp_pos - contains the positions
struct ggml_tensor * inp_pos = build_inp_pos();
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
for (int il = 0; il < n_layer; ++il) {
auto residual = inpL;
// self-attention
{
struct ggml_tensor* attn_norm_output = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm,
NULL,
LLM_NORM_RMS, cb, il);
cb(attn_norm_output, "attn_norm", il);
struct ggml_tensor * Qcur = nullptr;
struct ggml_tensor * Kcur = nullptr;
struct ggml_tensor * Vcur = nullptr;
if (model.layers[il].wqkv) {
cur = ggml_mul_mat(ctx0, model.layers[il].wqkv, attn_norm_output);
cb(cur, "wqkv", il);
Qcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd, n_tokens, cur->nb[1], 0 * sizeof(float) * (n_embd)));
Kcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1 * sizeof(float) * (n_embd)));
Vcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1 * sizeof(float) * (n_embd + n_embd_gqa)));
}
else {
Qcur = ggml_add(ctx0, ggml_mul_mat(ctx0, model.layers[il].wq, attn_norm_output), model.layers[il].bq);
Kcur = ggml_add(ctx0, ggml_mul_mat(ctx0, model.layers[il].wk, attn_norm_output), model.layers[il].bk);
Vcur = ggml_add(ctx0, ggml_mul_mat(ctx0, model.layers[il].wv, attn_norm_output), model.layers[il].bv);
}
cb(Qcur, "Qcur", il);
cb(Kcur, "Kcur", il);
cb(Vcur, "Vcur", il);
Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
Qcur = ggml_rope_custom(
ctx0, Qcur, inp_pos, n_rot, rope_type, 0, n_orig_ctx,
freq_base, freq_scale, ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Qcur, "Qcur", il);
Qcur = ggml_scale(ctx0, Qcur, 1.0f / sqrtf(float(n_embd_head)));
cb(Qcur, "Qcur", il);
Kcur = ggml_rope_custom(
ctx0, Kcur, inp_pos, n_rot, rope_type, 0, n_orig_ctx,
freq_base, freq_scale, ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
model.layers[il].wo, model.layers[il].bo,
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f, cb, il);
llama : add phi3 support (#6852) * add explicit phi3 support * add explicit phi3 support * remove unused code * convert : add BOS token * llama : match EOT token <|end|> * llama : minor / style * llama : tabs -> spaces * convert : fix lint checks --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-24 07:00:37 +00:00
}
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor* inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
residual = ggml_get_rows(ctx0, residual, inp_out_ids);
}
cur = ggml_add(ctx0, cur, residual);
residual = cur;
cur = llm_build_norm(ctx0, cur, hparams,
model.layers[il].ffn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "ffn_norm", il);
// FF
// special-case: the up and gate tensors are merged into a single tensor
// TOOD: support into llm_build_ffn
{
struct ggml_tensor* up = ggml_mul_mat(ctx0, model.layers[il].ffn_up, cur);
cb(up, "ffn_up", il);
auto g = ggml_cont(ctx0, ggml_view_2d(ctx0, up, up->ne[0] / 2, up->ne[1], ggml_row_size(up->type, up->ne[0]), 0));
auto y = ggml_cont(ctx0, ggml_view_2d(ctx0, up, up->ne[0] / 2, up->ne[1], ggml_row_size(up->type, up->ne[0]), up->nb[1] / 2));
y = ggml_mul(ctx0, y, ggml_silu(ctx0, g));
cb(y, "ffn_gate", il);
auto down = ggml_mul_mat(ctx0, model.layers[il].ffn_down, y);
cb(down, "ffn_down", il);
cur = down;
cb(cur, "ffn_out", il);
}
cur = ggml_add(ctx0, residual, cur);
cb(cur, "l_out", il);
inpL = cur;
}
cur = llm_build_norm(ctx0, inpL, hparams,
model.output_norm,
NULL,
LLM_NORM_RMS, cb, -1);
cb(cur, "result_norm", -1);
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
ggml_build_forward_expand(gf, cur);
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
return gf;
}
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
llama : add phi3 support (#6852) * add explicit phi3 support * add explicit phi3 support * remove unused code * convert : add BOS token * llama : match EOT token <|end|> * llama : minor / style * llama : tabs -> spaces * convert : fix lint checks --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-24 07:00:37 +00:00
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
struct ggml_cgraph * build_plamo() {
struct ggml_cgraph * gf = ggml_new_graph(ctx0);
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_embd_head = hparams.n_embd_head_v;
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
llama : fix llm_build_k_shift to use correct n_rot (#4889) * llama : fix llm_build_k_shift to use correct n_rot ggml-ci * llama : always use hparams.n_rot for ggml_rope_custom ggml-ci * convert : fix persimmon conversion to write correct n_rot
2024-01-12 11:01:56 +00:00
GGML_ASSERT(n_embd_head == hparams.n_rot);
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
// inp_pos - contains the positions
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * inp_pos = build_inp_pos();
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
for (int il = 0; il < n_layer; ++il) {
// norm
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "attn_norm", il);
struct ggml_tensor * attention_norm = cur;
// self-attention
{
// compute Q and K and RoPE them
struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
cb(Qcur, "Qcur", il);
struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur);
cb(Kcur, "Kcur", il);
struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur);
cb(Vcur, "Vcur", il);
Qcur = ggml_rope_custom(
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ctx0, ggml_reshape_3d(ctx0, Qcur, n_rot, n_head, n_tokens), inp_pos,
n_embd_head, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
ext_factor, attn_factor, beta_fast, beta_slow);
cb(Qcur, "Qcur", il);
Kcur = ggml_rope_custom(
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ctx0, ggml_reshape_3d(ctx0, Kcur, n_rot, n_head_kv, n_tokens), inp_pos,
n_embd_head, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
ext_factor, attn_factor, beta_fast, beta_slow);
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
model.layers[il].wo, NULL,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
}
struct ggml_tensor * sa_out = cur;
cur = attention_norm;
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
sa_out = ggml_get_rows(ctx0, sa_out, inp_out_ids);
inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
}
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
// feed-forward network
{
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, NULL,
model.layers[il].ffn_gate, NULL,
model.layers[il].ffn_down, NULL,
llama : add AWQ for llama, llama2, mpt, and mistral models (#4593) * update: awq support llama-7b model * update: change order * update: benchmark results for llama2-7b * update: mistral 7b v1 benchmark * update: support 4 models * fix: Readme * update: ready for PR * update: readme * fix: readme * update: change order import * black * format code * update: work for bot mpt and awqmpt * update: readme * Rename to llm_build_ffn_mpt_awq * Formatted other files * Fixed params count * fix: remove code * update: more detail for mpt * fix: readme * fix: readme * update: change folder architecture * fix: common.cpp * fix: readme * fix: remove ggml_repeat * update: cicd * update: cicd * uppdate: remove use_awq arg * update: readme * llama : adapt plamo to new ffn ggml-ci --------- Co-authored-by: Trần Đức Nam <v.namtd12@vinai.io> Co-authored-by: Le Hoang Anh <v.anhlh33@vinai.io> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-27 15:39:45 +00:00
NULL,
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
cb(cur, "ffn_out", il);
}
cur = ggml_add(ctx0, cur, sa_out);
cb(cur, "l_out", il);
cur = ggml_add(ctx0, cur, inpL);
cb(cur, "l_out", il);
// input for next layer
inpL = cur;
}
cur = inpL;
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm, NULL,
LLM_NORM_RMS, cb, -1);
cb(cur, "result_norm", -1);
// lm_head
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
return gf;
}
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
struct ggml_cgraph * build_gpt2() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
const int64_t n_embd_head = hparams.n_embd_head_v;
const int64_t n_embd_gqa = hparams.n_embd_v_gqa();
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
struct ggml_tensor * cur;
struct ggml_tensor * pos;
struct ggml_tensor * inpL;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
// inp_pos - contains the positions
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * inp_pos = build_inp_pos();
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
pos = ggml_get_rows(ctx0, model.pos_embd, inp_pos);
cb(pos, "pos_embd", -1);
inpL = ggml_add(ctx0, inpL, pos);
cb(inpL, "inpL", -1);
for (int il = 0; il < n_layer; ++il) {
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm,
model.layers[il].attn_norm_b,
LLM_NORM, cb, il);
cb(cur, "attn_norm", il);
// self-attention
{
cur = ggml_mul_mat(ctx0, model.layers[il].wqkv, cur);
cb(cur, "wqkv", il);
cur = ggml_add(ctx0, cur, model.layers[il].bqkv);
cb(cur, "bqkv", il);
struct ggml_tensor * Qcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd, n_tokens, cur->nb[1], 0*sizeof(float)*(n_embd)));
struct ggml_tensor * Kcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd)));
struct ggml_tensor * Vcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd + n_embd_gqa)));
cb(Qcur, "Qcur", il);
cb(Kcur, "Kcur", il);
cb(Vcur, "Vcur", il);
Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
model.layers[il].wo, model.layers[il].bo,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
}
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
// add the input
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL);
cb(ffn_inp, "ffn_inp", il);
// FF
{
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm,
model.layers[il].ffn_norm_b,
LLM_NORM, cb, il);
cb(cur, "ffn_norm", il);
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, model.layers[il].ffn_up_b,
NULL, NULL,
model.layers[il].ffn_down, model.layers[il].ffn_down_b,
NULL,
LLM_FFN_GELU, LLM_FFN_SEQ, cb, il);
cb(cur, "ffn_out", il);
}
inpL = ggml_add(ctx0, cur, ffn_inp);
cb(inpL, "l_out", il);
}
cur = llm_build_norm(ctx0, inpL, hparams,
model.output_norm,
model.output_norm_b,
LLM_NORM, cb, -1);
cb(cur, "result_norm", -1);
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
return gf;
}
llama : add CodeShell support (#5016) * llama: add codeshell support * llama.cpp: fix codeshell with NeoX rope Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-19 09:07:27 +00:00
struct ggml_cgraph * build_codeshell() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
const int64_t n_embd_head = hparams.n_embd_head_v;
const int64_t n_embd_gqa = hparams.n_embd_v_gqa();
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
GGML_ASSERT(n_embd_head == hparams.n_rot);
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llama : add CodeShell support (#5016) * llama: add codeshell support * llama.cpp: fix codeshell with NeoX rope Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-19 09:07:27 +00:00
// inp_pos - contains the positions
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * inp_pos = build_inp_pos();
llama : add CodeShell support (#5016) * llama: add codeshell support * llama.cpp: fix codeshell with NeoX rope Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-19 09:07:27 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llama : add CodeShell support (#5016) * llama: add codeshell support * llama.cpp: fix codeshell with NeoX rope Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-19 09:07:27 +00:00
for (int il = 0; il < n_layer; ++il) {
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm,
model.layers[il].attn_norm_b,
LLM_NORM, cb, il);
cb(cur, "attn_norm", il);
// self-attention
{
cur = ggml_mul_mat(ctx0, model.layers[il].wqkv, cur);
cb(cur, "wqkv", il);
cur = ggml_add(ctx0, cur, model.layers[il].bqkv);
cb(cur, "bqkv", il);
struct ggml_tensor * tmpq = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd, n_tokens, cur->nb[1], 0*sizeof(float)*(n_embd)));
struct ggml_tensor * tmpk = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd)));
struct ggml_tensor * Vcur = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, n_embd_gqa, n_tokens, cur->nb[1], 1*sizeof(float)*(n_embd + n_embd_gqa)));
cb(tmpq, "tmpq", il);
cb(tmpk, "tmpk", il);
cb(Vcur, "Vcur", il);
struct ggml_tensor * Qcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, tmpq, n_embd_head, n_head, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : add CodeShell support (#5016) * llama: add codeshell support * llama.cpp: fix codeshell with NeoX rope Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-19 09:07:27 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Qcur, "Qcur", il);
struct ggml_tensor * Kcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, tmpk, n_embd_head, n_head_kv, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : add CodeShell support (#5016) * llama: add codeshell support * llama.cpp: fix codeshell with NeoX rope Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-19 09:07:27 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llama : add CodeShell support (#5016) * llama: add codeshell support * llama.cpp: fix codeshell with NeoX rope Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-19 09:07:27 +00:00
model.layers[il].wo, model.layers[il].bo,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llama : add CodeShell support (#5016) * llama: add codeshell support * llama.cpp: fix codeshell with NeoX rope Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-19 09:07:27 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
}
llama : add CodeShell support (#5016) * llama: add codeshell support * llama.cpp: fix codeshell with NeoX rope Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-19 09:07:27 +00:00
// add the input
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpL);
cb(ffn_inp, "ffn_inp", il);
// FF
{
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm,
model.layers[il].ffn_norm_b,
LLM_NORM, cb, il);
cb(cur, "ffn_norm", il);
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, model.layers[il].ffn_up_b,
NULL, NULL,
model.layers[il].ffn_down, model.layers[il].ffn_down_b,
NULL,
LLM_FFN_GELU, LLM_FFN_SEQ, cb, il);
cb(cur, "ffn_out", il);
}
inpL = ggml_add(ctx0, cur, ffn_inp);
cb(inpL, "l_out", il);
}
cur = llm_build_norm(ctx0, inpL, hparams,
model.output_norm,
model.output_norm_b,
LLM_NORM, cb, -1);
cb(cur, "result_norm", -1);
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
return gf;
}
llama : reorder build_orion() at correct place (#5118)
2024-01-31 16:47:10 +00:00
struct ggml_cgraph * build_orion() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
const int64_t n_embd_head = hparams.n_embd_head_v;
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
GGML_ASSERT(n_embd_head == hparams.n_rot);
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llama : reorder build_orion() at correct place (#5118)
2024-01-31 16:47:10 +00:00
// inp_pos - contains the positions
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * inp_pos = build_inp_pos();
llama : reorder build_orion() at correct place (#5118)
2024-01-31 16:47:10 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llama : reorder build_orion() at correct place (#5118)
2024-01-31 16:47:10 +00:00
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * inpSA = inpL;
// norm
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm, model.layers[il].attn_norm_b,
LLM_NORM, cb, il);
cb(cur, "attn_norm", il);
// self-attention
{
// compute Q and K and RoPE them
struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
cb(Qcur, "Qcur", il);
// if (model.layers[il].bq) {
// Qcur = ggml_add(ctx0, Qcur, model.layers[il].bq);
// cb(Qcur, "Qcur", il);
// }
struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur);
cb(Kcur, "Kcur", il);
// if (model.layers[il].bk) {
// Kcur = ggml_add(ctx0, Kcur, model.layers[il].bk);
// cb(Kcur, "Kcur", il);
// }
struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur);
cb(Vcur, "Vcur", il);
// if (model.layers[il].bv) {
// Vcur = ggml_add(ctx0, Vcur, model.layers[il].bv);
// cb(Vcur, "Vcur", il);
// }
Qcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : reorder build_orion() at correct place (#5118)
2024-01-31 16:47:10 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Qcur, "Qcur", il);
Kcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : reorder build_orion() at correct place (#5118)
2024-01-31 16:47:10 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llama : reorder build_orion() at correct place (#5118)
2024-01-31 16:47:10 +00:00
model.layers[il].wo, NULL,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llama : reorder build_orion() at correct place (#5118)
2024-01-31 16:47:10 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
}
llama : reorder build_orion() at correct place (#5118)
2024-01-31 16:47:10 +00:00
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
cb(ffn_inp, "ffn_inp", il);
// feed-forward network
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm, model.layers[il].ffn_norm_b,
LLM_NORM, cb, il);
cb(cur, "ffn_norm", il);
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, NULL,
model.layers[il].ffn_gate, NULL,
model.layers[il].ffn_down, NULL,
NULL,
LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
cb(cur, "ffn_out", il);
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "l_out", il);
// input for next layer
inpL = cur;
}
cur = inpL;
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm, model.output_norm_b,
LLM_NORM, cb, -1);
cb(cur, "result_norm", -1);
// lm_head
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
return gf;
}
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
struct ggml_cgraph * build_internlm2() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
const int64_t n_embd_head = hparams.n_embd_head_v;
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
GGML_ASSERT(n_embd_head == hparams.n_rot);
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
// inp_pos - contains the positions
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * inp_pos = build_inp_pos();
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * inpSA = inpL;
// norm
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "attn_norm", il);
// self-attention
{
// compute Q and K and RoPE them
struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
cb(Qcur, "Qcur", il);
if (model.layers[il].bq) {
Qcur = ggml_add(ctx0, Qcur, model.layers[il].bq);
cb(Qcur, "Qcur", il);
}
struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur);
cb(Kcur, "Kcur", il);
if (model.layers[il].bk) {
Kcur = ggml_add(ctx0, Kcur, model.layers[il].bk);
cb(Kcur, "Kcur", il);
}
struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur);
cb(Vcur, "Vcur", il);
if (model.layers[il].bv) {
Vcur = ggml_add(ctx0, Vcur, model.layers[il].bv);
cb(Vcur, "Vcur", il);
}
Qcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Qcur, "Qcur", il);
Kcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
model.layers[il].wo, model.layers[il].bo,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
}
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
cb(ffn_inp, "ffn_inp", il);
// feed-forward network
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "ffn_norm", il);
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, NULL,
model.layers[il].ffn_gate, NULL,
model.layers[il].ffn_down, NULL,
NULL,
LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
cb(cur, "ffn_out", il);
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "l_out", il);
// input for next layer
inpL = cur;
}
cur = inpL;
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm, NULL,
LLM_NORM_RMS, cb, -1);
cb(cur, "result_norm", -1);
// lm_head
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
return gf;
}
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
// ref: https://arxiv.org/abs/2203.03466
// https://github.com/ggerganov/llama.cpp/issues/5276#issuecomment-1925774738
// based on the original build_llama() function
struct ggml_cgraph * build_minicpm() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
const int64_t n_embd_head = hparams.n_embd_head_v;
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
GGML_ASSERT(n_embd_head == hparams.n_rot);
const int64_t n_embd = hparams.n_embd;
//TODO: if the model varies, these parameters need to be read from the model
const int64_t n_embd_base = 256;
const float scale_embd = 12.0f;
const float scale_depth = 1.4f;
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
// scale the input embeddings
inpL = ggml_scale(ctx0, inpL, scale_embd);
cb(inpL, "inp_scaled", -1);
// inp_pos - contains the positions
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * inp_pos = build_inp_pos();
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * inpSA = inpL;
// norm
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "attn_norm", il);
// self-attention
{
// compute Q and K and RoPE them
struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
cb(Qcur, "Qcur", il);
if (model.layers[il].bq) {
Qcur = ggml_add(ctx0, Qcur, model.layers[il].bq);
cb(Qcur, "Qcur", il);
}
struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur);
cb(Kcur, "Kcur", il);
if (model.layers[il].bk) {
Kcur = ggml_add(ctx0, Kcur, model.layers[il].bk);
cb(Kcur, "Kcur", il);
}
struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur);
cb(Vcur, "Vcur", il);
if (model.layers[il].bv) {
Vcur = ggml_add(ctx0, Vcur, model.layers[il].bv);
cb(Vcur, "Vcur", il);
}
Qcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Qcur, "Qcur", il);
Kcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
model.layers[il].wo, model.layers[il].bo,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
}
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
// scale_res - scale the hidden states for residual connection
const float scale_res = scale_depth/sqrtf(float(n_layer));
cur = ggml_scale(ctx0, cur, scale_res);
cb(cur, "hidden_scaled", -1);
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
cb(ffn_inp, "ffn_inp", il);
// feed-forward network
{
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "ffn_norm", il);
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, NULL,
model.layers[il].ffn_gate, NULL,
model.layers[il].ffn_down, NULL,
NULL,
LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
cb(cur, "ffn_out", il);
}
// scale the hidden states for residual connection
cur = ggml_scale(ctx0, cur, scale_res);
cb(cur, "hidden_scaled_ffn", -1);
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "l_out", il);
// input for next layer
inpL = cur;
}
cur = inpL;
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm, NULL,
LLM_NORM_RMS, cb, -1);
cb(cur, "result_norm", -1);
// lm_head scaling
const float scale_lmhead = float(n_embd_base)/float(n_embd);
cur = ggml_scale(ctx0, cur, scale_lmhead);
cb(cur, "lmhead_scaling", -1);
// lm_head
cur = ggml_mul_mat(ctx0, model.tok_embd, cur);
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
return gf;
}
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
struct ggml_cgraph * build_gemma() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
const int64_t n_embd_head_k = hparams.n_embd_head_k;
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
py : add Gemma conversion from HF models (#5647) * py : add gemma conversion from HF models * Update convert-hf-to-gguf.py Co-authored-by: Aarni Koskela <akx@iki.fi> * Update convert-hf-to-gguf.py Co-authored-by: Aarni Koskela <akx@iki.fi> * Update convert-hf-to-gguf.py Co-authored-by: Jared Van Bortel <jared@nomic.ai> --------- Co-authored-by: Aarni Koskela <akx@iki.fi> Co-authored-by: Jared Van Bortel <jared@nomic.ai>
2024-02-22 21:22:48 +00:00
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
inpL = ggml_scale(ctx0, inpL, sqrtf(n_embd));
cb(inpL, "inp_scaled", -1);
// inp_pos - contains the positions
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * inp_pos = build_inp_pos();
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
for (int il = 0; il < n_layer; ++il) {
// norm
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "attn_norm", il);
// self-attention
{
// compute Q and K and RoPE them
struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
cb(Qcur, "Qcur", il);
struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur);
cb(Kcur, "Kcur", il);
struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur);
cb(Vcur, "Vcur", il);
Qcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head_k, n_head, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_embd_head_k, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
ext_factor, attn_factor, beta_fast, beta_slow);
cb(Qcur, "Qcur", il);
py : add Gemma conversion from HF models (#5647) * py : add gemma conversion from HF models * Update convert-hf-to-gguf.py Co-authored-by: Aarni Koskela <akx@iki.fi> * Update convert-hf-to-gguf.py Co-authored-by: Aarni Koskela <akx@iki.fi> * Update convert-hf-to-gguf.py Co-authored-by: Jared Van Bortel <jared@nomic.ai> --------- Co-authored-by: Aarni Koskela <akx@iki.fi> Co-authored-by: Jared Van Bortel <jared@nomic.ai>
2024-02-22 21:22:48 +00:00
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
Qcur = ggml_scale(ctx0, Qcur, 1.0f / sqrtf(float(n_embd_head_k)));
cb(Qcur, "Qcur_scaled", il);
Kcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head_k, n_head_kv, n_tokens), inp_pos,
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
n_embd_head_k, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
ext_factor, attn_factor, beta_fast, beta_slow);
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
model.layers[il].wo, NULL,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f, cb, il);
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
}
py : add Gemma conversion from HF models (#5647) * py : add gemma conversion from HF models * Update convert-hf-to-gguf.py Co-authored-by: Aarni Koskela <akx@iki.fi> * Update convert-hf-to-gguf.py Co-authored-by: Aarni Koskela <akx@iki.fi> * Update convert-hf-to-gguf.py Co-authored-by: Jared Van Bortel <jared@nomic.ai> --------- Co-authored-by: Aarni Koskela <akx@iki.fi> Co-authored-by: Jared Van Bortel <jared@nomic.ai>
2024-02-22 21:22:48 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
}
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
struct ggml_tensor * sa_out = ggml_add(ctx0, cur, inpL);
cb(sa_out, "sa_out", il);
cur = llm_build_norm(ctx0, sa_out, hparams,
model.layers[il].ffn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "ffn_norm", il);
// feed-forward network
{
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, NULL,
model.layers[il].ffn_gate, NULL,
model.layers[il].ffn_down, NULL,
NULL,
LLM_FFN_GELU, LLM_FFN_PAR, cb, il);
cb(cur, "ffn_out", il);
}
cur = ggml_add(ctx0, cur, sa_out);
cb(cur, "l_out", il);
// input for next layer
inpL = cur;
}
cur = inpL;
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm, NULL,
LLM_NORM_RMS, cb, -1);
cb(cur, "result_norm", -1);
// lm_head
gemma : allow offloading the output tensor (#5646)
2024-02-21 21:18:23 +00:00
cur = ggml_mul_mat(ctx0, model.output, cur);
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
return gf;
}
llama : add StarCoder2 support (#5795) * Add support for starcoder2 * handle rope type * skip rope freq and rotary embeddings from being serialized * resolve comments * Update llama.cpp * remove redundant changes * handle `rope-theta` * llama : change starcoder2 rope type * address comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-01 19:30:46 +00:00
struct ggml_cgraph * build_starcoder2() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
const int64_t n_embd_head = hparams.n_embd_head_v;
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
GGML_ASSERT(n_embd_head == hparams.n_rot);
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llama : add StarCoder2 support (#5795) * Add support for starcoder2 * handle rope type * skip rope freq and rotary embeddings from being serialized * resolve comments * Update llama.cpp * remove redundant changes * handle `rope-theta` * llama : change starcoder2 rope type * address comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-01 19:30:46 +00:00
// inp_pos - contains the positions
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * inp_pos = build_inp_pos();
llama : add StarCoder2 support (#5795) * Add support for starcoder2 * handle rope type * skip rope freq and rotary embeddings from being serialized * resolve comments * Update llama.cpp * remove redundant changes * handle `rope-theta` * llama : change starcoder2 rope type * address comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-01 19:30:46 +00:00
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
llama : add StarCoder2 support (#5795) * Add support for starcoder2 * handle rope type * skip rope freq and rotary embeddings from being serialized * resolve comments * Update llama.cpp * remove redundant changes * handle `rope-theta` * llama : change starcoder2 rope type * address comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-01 19:30:46 +00:00
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * inpSA = inpL;
// norm
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm, model.layers[il].attn_norm_b,
LLM_NORM, cb, il);
cb(cur, "attn_norm", il);
// self-attention
{
// compute Q and K and RoPE them
struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
cb(Qcur, "Qcur", il);
if (model.layers[il].bq) {
Qcur = ggml_add(ctx0, Qcur, model.layers[il].bq);
cb(Qcur, "Qcur", il);
}
struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur);
cb(Kcur, "Kcur", il);
if (model.layers[il].bk) {
Kcur = ggml_add(ctx0, Kcur, model.layers[il].bk);
cb(Kcur, "Kcur", il);
}
struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur);
cb(Vcur, "Vcur", il);
if (model.layers[il].bv) {
Vcur = ggml_add(ctx0, Vcur, model.layers[il].bv);
cb(Vcur, "Vcur", il);
}
Qcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens), inp_pos,
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Qcur, "Qcur", il);
Kcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens), inp_pos,
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llama : add StarCoder2 support (#5795) * Add support for starcoder2 * handle rope type * skip rope freq and rotary embeddings from being serialized * resolve comments * Update llama.cpp * remove redundant changes * handle `rope-theta` * llama : change starcoder2 rope type * address comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-01 19:30:46 +00:00
model.layers[il].wo, model.layers[il].bo,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llama : add StarCoder2 support (#5795) * Add support for starcoder2 * handle rope type * skip rope freq and rotary embeddings from being serialized * resolve comments * Update llama.cpp * remove redundant changes * handle `rope-theta` * llama : change starcoder2 rope type * address comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-01 19:30:46 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
}
llama : add StarCoder2 support (#5795) * Add support for starcoder2 * handle rope type * skip rope freq and rotary embeddings from being serialized * resolve comments * Update llama.cpp * remove redundant changes * handle `rope-theta` * llama : change starcoder2 rope type * address comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-01 19:30:46 +00:00
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
cb(ffn_inp, "ffn_inp", il);
// feed-forward network
cur = llm_build_norm(ctx0, ffn_inp, hparams,
model.layers[il].ffn_norm, model.layers[il].ffn_norm_b,
LLM_NORM, cb, il);
cb(cur, "ffn_norm", il);
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, model.layers[il].ffn_up_b,
NULL, NULL,
model.layers[il].ffn_down, model.layers[il].ffn_down_b,
NULL,
LLM_FFN_GELU, LLM_FFN_SEQ, cb, il);
cb(cur, "ffn_out", il);
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "l_out", il);
// input for next layer
inpL = cur;
}
cur = inpL;
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm, model.output_norm_b,
LLM_NORM, cb, -1);
cb(cur, "result_norm", -1);
// lm_head
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
return gf;
}
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
struct ggml_cgraph * build_mamba() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
const int64_t d_model = n_embd;
const int64_t d_conv = hparams.ssm_d_conv;
const int64_t d_inner = hparams.ssm_d_inner;
GGML_ASSERT(2 * d_model == d_inner);
const int64_t d_state = hparams.ssm_d_state;
const int64_t dt_rank = hparams.ssm_dt_rank;
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
// {n_embd, n_tokens}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
struct ggml_tensor * state_mask = build_inp_s_mask();
struct ggml_tensor * state_seq = build_inp_s_seq();
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
for (int il = 0; il < n_layer; ++il) {
// (ab)using the KV cache to store the states
struct ggml_tensor * conv_states = ggml_reshape_2d(ctx0, kv_self.k_l[il], hparams.n_embd_k_s(), kv_self.size);
struct ggml_tensor * ssm_states = ggml_reshape_2d(ctx0, kv_self.v_l[il], hparams.n_embd_v_s(), kv_self.size);
// clear states of sequences which are starting at the beginning of this batch
{
conv_states = ggml_mul(ctx0,
ggml_view_2d(ctx0, conv_states, conv_states->ne[0], n_kv, conv_states->nb[1], kv_head*conv_states->nb[1]),
state_mask);
ssm_states = ggml_mul(ctx0,
ggml_view_2d(ctx0, ssm_states, ssm_states->ne[0], n_kv, ssm_states->nb[1], kv_head*ssm_states->nb[1]),
state_mask);
}
conv_states = ggml_reshape_3d(ctx0, conv_states, d_conv - 1, d_inner, n_kv);
ssm_states = ggml_reshape_3d(ctx0, ssm_states, d_state, d_inner, n_kv);
// norm
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm, NULL,
LLM_NORM_RMS, cb, il);
cb(cur, "attn_norm", il);
// {n_embd, 2*d_inner} * {n_embd, n_tokens} => {2*d_inner, n_tokens}
struct ggml_tensor * xz = ggml_mul_mat(ctx0, model.layers[il].ssm_in, cur);
// split the above in two
// => {d_inner, n_tokens}
struct ggml_tensor * x = ggml_view_2d(ctx0, xz, d_inner, xz->ne[1], xz->nb[1], 0);
struct ggml_tensor * z = ggml_view_2d(ctx0, xz, d_inner, xz->ne[1], xz->nb[1], ggml_element_size(xz)*d_inner);
// conv
{
// Custom operator which is needed only to ease simultaneous sequence processing.
// For a single sequence, the equivalent is to concatenate the columns of conv_states and x,
// then make a self-overlapping view of that over d_conv columns at each stride in the 3rd dimension,
// then element-wise multiply that with the conv1d weigth,
// then sum the elements of each row,
// (the last two steps are a dot product over rows (also doable with mul_mat))
// then permute away the ne[0] dimension,
// and then you're left with the resulting x tensor.
// The new conv_states is the last (d_conv - 1) columns
// of the last 3rd dimensional "layer" of the self-overlapping view.
// For simultaneous sequences, it's more complicated.
struct ggml_tensor * x_conv = ggml_ssm_conv(ctx0, conv_states, x, model.layers[il].ssm_conv1d, state_seq);
// store last (d_conv - 1) columns of the conv_state part of x_conv back into the KV cache
ggml_build_forward_expand(gf,
ggml_cpy(ctx0,
ggml_view_2d(ctx0, x_conv, d_conv - 1, d_inner*n_kv, d_conv*ggml_element_size(x_conv), (1+d_inner*n_tokens)*ggml_element_size(x_conv)),
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
ggml_view_1d(ctx0, kv_self.k_l[il], (d_conv - 1)*(d_inner)*(n_kv), kv_head*(d_conv - 1)*(d_inner)*ggml_element_size(x_conv))));
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
// extract x from x_conv
x = ggml_view_2d(ctx0, x_conv, d_inner, n_tokens, d_inner*ggml_element_size(x_conv), 0);
// bias
x = ggml_add(ctx0, x, model.layers[il].ssm_conv1d_b);
x = ggml_silu(ctx0, x);
}
// ssm
{
// {d_inner, dt_rank + 2*d_state} * {d_inner, n_tokens} => {dt_rank + 2*d_state, n_tokens}
struct ggml_tensor * x_db = ggml_mul_mat(ctx0, model.layers[il].ssm_x, x);
// split
struct ggml_tensor * dt = ggml_view_2d(ctx0, x_db, dt_rank, n_tokens, x_db->nb[1], 0);
struct ggml_tensor * B = ggml_view_2d(ctx0, x_db, d_state, n_tokens, x_db->nb[1], ggml_element_size(x_db)*dt_rank);
struct ggml_tensor * C = ggml_view_2d(ctx0, x_db, d_state, n_tokens, x_db->nb[1], ggml_element_size(x_db)*(dt_rank+d_state));
// {dt_rank, d_inner} * {dt_rank, n_tokens} => {d_inner, n_tokens}
dt = ggml_mul_mat(ctx0, model.layers[il].ssm_dt, dt);
dt = ggml_add(ctx0, dt, model.layers[il].ssm_dt_b);
// Custom operator to optimize the parallel associative scan
// as described in the Annex D of the Mamba paper.
// => {d_inner, n_tokens} and {d_state, d_inner, n_kv} combined,
// because only a single tensor can be returned.
struct ggml_tensor * y_ssm_states = ggml_ssm_scan(ctx0, ssm_states, x, dt, model.layers[il].ssm_a, B, C, state_seq);
// store last states (the second part of y_ssm_states)
ggml_build_forward_expand(gf,
ggml_cpy(ctx0,
ggml_view_1d(ctx0, y_ssm_states, d_state*d_inner*n_kv, d_inner*n_tokens*ggml_element_size(y_ssm_states)),
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
ggml_view_1d(ctx0, kv_self.v_l[il], d_state*d_inner*n_kv, kv_head*d_state*d_inner*ggml_element_size(ssm_states))));
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
struct ggml_tensor * y = ggml_view_2d(ctx0, y_ssm_states, d_inner, n_tokens, d_inner*ggml_element_size(y_ssm_states), 0);
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
x = ggml_get_rows(ctx0, x, inp_out_ids);
y = ggml_get_rows(ctx0, y, inp_out_ids);
z = ggml_get_rows(ctx0, z, inp_out_ids);
inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
}
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
// {d_inner, n_tokens} * {d_inner} => {d_inner, n_tokens}
y = ggml_add(ctx0, y, ggml_mul(ctx0, x, model.layers[il].ssm_d));
y = ggml_mul(ctx0, y, ggml_silu(ctx0, z));
// {d_inner, n_embd} * {d_inner, n_tokens} => {n_embd, n_tokens}
cur = ggml_mul_mat(ctx0, model.layers[il].ssm_out, y);
}
// residual
cur = ggml_add(ctx0, cur, inpL);
cb(cur, "l_out", il);
// input for next layer
inpL = cur;
}
// final rmsnorm
cur = llm_build_norm(ctx0, inpL, hparams,
model.output_norm, NULL,
LLM_NORM_RMS, cb, -1);
cb(cur, "result_norm", -1);
// lm_head
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
return gf;
}
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
struct ggml_cgraph * build_command_r() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
const int64_t n_embd_head = hparams.n_embd_head_v;
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
const float f_logit_scale = hparams.f_logit_scale;
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
// inp_pos - contains the positions
struct ggml_tensor * inp_pos = build_inp_pos();
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
for (int il = 0; il < n_layer; ++il) {
// norm
cur = llm_build_norm(ctx0, inpL, hparams,
model.layers[il].attn_norm, NULL,
LLM_NORM, cb, il);
cb(cur, "attn_norm", il);
struct ggml_tensor * ffn_inp = cur;
// self-attention
{
// compute Q and K and RoPE them
struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
cb(Qcur, "Qcur", il);
if (model.layers[il].bq) {
Qcur = ggml_add(ctx0, Qcur, model.layers[il].bq);
cb(Qcur, "Qcur", il);
}
struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur);
cb(Kcur, "Kcur", il);
if (model.layers[il].bk) {
Kcur = ggml_add(ctx0, Kcur, model.layers[il].bk);
cb(Kcur, "Kcur", il);
}
struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur);
cb(Vcur, "Vcur", il);
if (model.layers[il].bv) {
Vcur = ggml_add(ctx0, Vcur, model.layers[il].bv);
cb(Vcur, "Vcur", il);
}
llama : add Command R Plus support (#6491) * Add Command R Plus GGUF * Add Command R Plus GGUF * Loading works up to LayerNorm2D * Export new tensors in 1D so they are not quantized. * Fix embedding layer based on Noeda's example * Whitespace * Add line * Fix unexpected tokens on MPS. Re-add F16 fix. ((Noeda) * dranger003: Fix block index overflow in CUDA dequantizing. * Reverted blocked multiplication code as it still has issues and could affect other Llama arches * export norms as f32 * fix overflow issues during quant and other cleanup * Type convention Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * dranger003: Fix more int overflow during quant. --------- Co-authored-by: S <seast@Ss-Mac-Studio.local> Co-authored-by: S <s@example.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-09 08:16:13 +00:00
if (model.layers[il].attn_q_norm) {
Qcur = ggml_view_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens,
ggml_element_size(Qcur) * n_embd_head,
ggml_element_size(Qcur) * n_embd_head * n_head,
0);
cb(Qcur, "Qcur", il);
Kcur = ggml_view_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens,
ggml_element_size(Kcur) * n_embd_head,
ggml_element_size(Kcur) * n_embd_head * n_head_kv,
0);
cb(Kcur, "Kcur", il);
Qcur = llm_build_norm(ctx0, Qcur, hparams,
model.layers[il].attn_q_norm,
NULL,
LLM_NORM, cb, il);
cb(Qcur, "Qcur", il);
Kcur = llm_build_norm(ctx0, Kcur, hparams,
model.layers[il].attn_k_norm,
NULL,
LLM_NORM, cb, il);
cb(Kcur, "Kcur", il);
}
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
Qcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens), inp_pos,
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Qcur, "Qcur", il);
Kcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens), inp_pos,
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
model.layers[il].wo, model.layers[il].bo,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
llama : fix command-r inference when omitting outputs (#6367)
2024-03-28 12:05:54 +00:00
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpL = ggml_get_rows(ctx0, inpL, inp_out_ids);
ffn_inp = ggml_get_rows(ctx0, ffn_inp, inp_out_ids);
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
}
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
struct ggml_tensor * attn_out = cur;
// feed-forward network
{
cur = llm_build_ffn(ctx0, ffn_inp,
model.layers[il].ffn_up, NULL,
model.layers[il].ffn_gate, NULL,
model.layers[il].ffn_down, NULL,
NULL,
LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
cb(cur, "ffn_out", il);
}
// add together residual + FFN + self-attention
cur = ggml_add(ctx0, cur, inpL);
cur = ggml_add(ctx0, cur, attn_out);
cb(cur, "l_out", il);
// input for next layer
inpL = cur;
}
cur = inpL;
cur = llm_build_norm(ctx0, cur, hparams,
model.output_norm, NULL,
LLM_NORM, cb, -1);
cb(cur, "result_norm", -1);
// lm_head
cur = ggml_mul_mat(ctx0, model.output, cur);
if (f_logit_scale) {
cur = ggml_scale(ctx0, cur, f_logit_scale);
}
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
return gf;
}
Implement the OLMo architecture (#6741) * implement olmo architecture * remove unused variable * remove unused moe branch * remove check for weight * remove superfluous moe, bias and rope tensors * clarified comment * fix clamp_kqv setting * remove obsolete parameter name filter
2024-04-19 09:35:54 +00:00
// ref: https://allenai.org/olmo
// based on the original build_llama() function, changes:
// * non-parametric layer norm
// * clamp qkv
// * removed bias
// * removed MoE
struct ggml_cgraph * build_olmo() {
struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false);
// mutable variable, needed during the last layer of the computation to skip unused tokens
int32_t n_tokens = this->n_tokens;
const int64_t n_embd_head = hparams.n_embd_head_v;
GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
GGML_ASSERT(n_embd_head == hparams.n_rot);
struct ggml_tensor * cur;
struct ggml_tensor * inpL;
inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb);
// inp_pos - contains the positions
struct ggml_tensor * inp_pos = build_inp_pos();
// KQ_mask (mask for 1 head, it will be broadcasted to all heads)
struct ggml_tensor * KQ_mask = build_inp_KQ_mask();
for (int il = 0; il < n_layer; ++il) {
struct ggml_tensor * inpSA = inpL;
// norm
cur = llm_build_norm(ctx0, inpL, hparams,
NULL, NULL,
LLM_NORM, cb, il);
cb(cur, "attn_norm", il);
// self-attention
{
// compute Q and K and RoPE them
struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur);
cb(Qcur, "Qcur", il);
if (hparams.f_clamp_kqv > 0.0f) {
Qcur = ggml_clamp(ctx0, Qcur, -hparams.f_clamp_kqv, hparams.f_clamp_kqv);
cb(Qcur, "Qcur", il);
}
struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur);
cb(Kcur, "Kcur", il);
if (hparams.f_clamp_kqv > 0.0f) {
Kcur = ggml_clamp(ctx0, Kcur, -hparams.f_clamp_kqv, hparams.f_clamp_kqv);
cb(Kcur, "Kcur", il);
}
struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur);
cb(Vcur, "Vcur", il);
if (hparams.f_clamp_kqv > 0.0f) {
Vcur = ggml_clamp(ctx0, Vcur, -hparams.f_clamp_kqv, hparams.f_clamp_kqv);
cb(Vcur, "Vcur", il);
}
Qcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens), inp_pos,
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Qcur, "Qcur", il);
Kcur = ggml_rope_custom(
ctx0, ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens), inp_pos,
n_rot, rope_type, 0, n_orig_ctx, freq_base, freq_scale,
ext_factor, attn_factor, beta_fast, beta_slow
);
cb(Kcur, "Kcur", il);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf,
Implement the OLMo architecture (#6741) * implement olmo architecture * remove unused variable * remove unused moe branch * remove check for weight * remove superfluous moe, bias and rope tensors * clarified comment * fix clamp_kqv setting * remove obsolete parameter name filter
2024-04-19 09:35:54 +00:00
model.layers[il].wo, nullptr,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
Kcur, Vcur, Qcur, KQ_mask, nullptr, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il);
Implement the OLMo architecture (#6741) * implement olmo architecture * remove unused variable * remove unused moe branch * remove check for weight * remove superfluous moe, bias and rope tensors * clarified comment * fix clamp_kqv setting * remove obsolete parameter name filter
2024-04-19 09:35:54 +00:00
}
if (il == n_layer - 1) {
// skip computing output for unused tokens
struct ggml_tensor * inp_out_ids = build_inp_out_ids();
n_tokens = n_outputs;
cur = ggml_get_rows(ctx0, cur, inp_out_ids);
inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
}
struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
cb(ffn_inp, "ffn_inp", il);
// feed-forward network
cur = llm_build_norm(ctx0, ffn_inp, hparams,
NULL, NULL,
LLM_NORM, cb, il);
cb(cur, "ffn_norm", il);
cur = llm_build_ffn(ctx0, cur,
model.layers[il].ffn_up, NULL,
model.layers[il].ffn_gate, NULL,
model.layers[il].ffn_down, NULL,
NULL,
LLM_FFN_SILU, LLM_FFN_PAR, cb, il);
cb(cur, "ffn_out", il);
cur = ggml_add(ctx0, cur, ffn_inp);
cb(cur, "ffn_out", il);
ggml_tensor * layer_dir = lctx.cvec.tensor_for(il);
if (layer_dir != nullptr) {
cur = ggml_add(ctx0, cur, layer_dir);
}
cb(cur, "l_out", il);
// input for next layer
inpL = cur;
}
cur = inpL;
cur = llm_build_norm(ctx0, cur, hparams,
NULL, NULL,
LLM_NORM, cb, -1);
cb(cur, "result_norm", -1);
// lm_head
cur = ggml_mul_mat(ctx0, model.output, cur);
cb(cur, "result_output", -1);
ggml_build_forward_expand(gf, cur);
return gf;
}
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
};
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
static struct ggml_cgraph * llama_build_graph_defrag(llama_context & lctx, const std::vector<uint32_t> & ids) {
llama_batch dummy;
dummy.n_tokens = 0;
llm_build_cb cb = [&](struct ggml_tensor * , const char * , int ) { };
struct llm_build_context llm(lctx, dummy, cb, false);
llm.init();
struct ggml_cgraph * result = llm.build_defrag(ids);
llm.free();
return result;
}
static struct ggml_cgraph * llama_build_graph_k_shift(llama_context & lctx) {
llama_batch dummy;
dummy.n_tokens = 0;
llm_build_cb cb = [&](struct ggml_tensor * , const char * , int ) { };
struct llm_build_context llm(lctx, dummy, cb, false);
llm.init();
struct ggml_cgraph * result = llm.build_k_shift();
llm.free();
return result;
}
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
static struct ggml_cgraph * llama_build_graph_s_copy(llama_context & lctx) {
llama_batch dummy;
dummy.n_tokens = 0;
llm_build_cb cb = [&](struct ggml_tensor * , const char * , int ) { };
struct llm_build_context llm(lctx, dummy, cb, false);
llm.init();
struct ggml_cgraph * result = llm.build_s_copy();
llm.free();
return result;
}
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
static struct ggml_cgraph * llama_build_graph(
llama_context & lctx,
sync : ggml (#5452) * ggml-alloc : v3 (ggml/727) * ggml-alloc v3 ggml-ci * fix ci ggml-ci * whisper : check for backend buffer allocation failures * whisper : avoid leaks when initialization fails * cleanup ggml-ci * style fixes ggml-ci * sync : ggml * update llama.cpp, clip.cpp, export-lora.cpp * update finetune.cpp, train-text-from-scratch.cpp ggml-ci * ggml-backend : reduce alignment to 32 to match gguf and fix mmap --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-12 07:16:06 +00:00
const llama_batch & batch,
bool worst_case) {
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
const auto & model = lctx.model;
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
llama : refactor graph build code (#3837) * llama : factor out ggml-alloc from graph graph build functions ggml-ci * metal : disable kernel load log * llama : factor out tensor offloading outside the build call (wip) ggml-ci * llama : offload rest of the models ggml-ci * llama : update offload log messages to print node index * llama : comments * llama : support offloading result_norm + comments * llama : factor graph input into a function * llama : do tensor offload only with CUDA * llama : fix res_norm offloading * llama : try to optimize offloading code * llama : fix non-CUDA build * llama : try to fix build * llama : move refact in correct place + optimize graph input * llama : refactor tensor offloading as callback * llama : add layer index to all tensor names * llama : add functional header * llama : comment ggml-ci * llama : remove obsolete map for layer counting * llama : add llm_build helper functions (#3848) * llama : add llm_build_norm helper function ggml-ci * llama : add llm_build_ffn helper function (#3849) ggml-ci * llama : add llm_build_k_shift helper ggml-ci * llama : fix offloading after recent changes * llama : add llm_build_kv_store helper ggml-ci * llama : remove obsolete offload names * llama : fix llm_build_k_shift to use n_head_kv instead of n_head * llama : simplify falcon Q, K, V computation * llama : remove obsolete comments in build graphs * llama : add llm_build_kqv helper ggml-ci * llama : minor * llama : add LLAMA_OFFLOAD_DEBUG + fix starcoder offloading * llama : fix input allocation logic * llama : update offload functions for KQ tensors * llama : normalize tensor names ggml-ci * llama : enable warning about not offloaded tensors * llama : remove extra ; + deduplicate gate_b logic * llama : add llm_build_inp_embd helper
2023-11-01 06:04:02 +00:00
// this callback allows us to apply custom logic to each tensor (e.g. ggml-alloc, offloading, etc.)
llm_build_cb cb = [&](struct ggml_tensor * cur, const char * name, int il) {
if (il >= 0) {
ggml_format_name(cur, "%s-%d", name, il);
} else {
ggml_set_name(cur, name);
}
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
llama : run all KQV ops on the CPU with no KV offload (#5049) ggml-ci
2024-01-20 15:05:49 +00:00
if (!lctx.cparams.offload_kqv) {
if (strcmp(name, "kqv_merged_cont") == 0) {
// all nodes between the KV store and the attention output are run on the CPU
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
ggml_backend_sched_set_tensor_backend(lctx.sched, cur, lctx.backend_cpu);
}
}
// norm may be automatically assigned to the backend of the previous layer, increasing data transfer between backends
backend : offload large batches to GPU (#6083) * backend : offload large batches to GPU * fix hip * code cleanup * fix CUDA split buffers * Update ggml-backend-impl.h Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix memset without set_device * imatrix : remove sched affix from weight names * sched : add a new split if the current one has too many inputs reduce max inputs per split more cleanup * update backends ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-03-18 10:03:04 +00:00
// FIXME: fix in ggml_backend_sched
const bool full_offload = lctx.model.n_gpu_layers > (int)lctx.model.hparams.n_layer;
if (batch.n_tokens < 32 || full_offload) {
if (il != -1 && strcmp(name, "norm") == 0) {
for (auto * backend : lctx.backends) {
if (ggml_backend_buft_supports_backend(lctx.model.buft_layer[il].buft, backend)) {
ggml_backend_sched_set_tensor_backend(lctx.sched, cur, backend);
break;
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
}
llama : run all KQV ops on the CPU with no KV offload (#5049) ggml-ci
2024-01-20 15:05:49 +00:00
}
}
llama : pre-allocate input tensors in a separate buffer (#5100)
2024-01-24 11:48:14 +00:00
};
llama : run all KQV ops on the CPU with no KV offload (#5049) ggml-ci
2024-01-20 15:05:49 +00:00
llama : pre-allocate input tensors in a separate buffer (#5100)
2024-01-24 11:48:14 +00:00
struct ggml_cgraph * result = NULL;
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
llama : pre-allocate input tensors in a separate buffer (#5100)
2024-01-24 11:48:14 +00:00
struct llm_build_context llm(lctx, batch, cb, worst_case);
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
llm.init();
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
switch (model.arch) {
case LLM_ARCH_LLAMA:
{
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
result = llm.build_llama();
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
} break;
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
case LLM_ARCH_BAICHUAN:
{
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
result = llm.build_baichuan();
feature : support Baichuan serial models (#3009)
2023-09-14 16:32:10 +00:00
} break;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
case LLM_ARCH_FALCON:
{
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
result = llm.build_falcon();
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
} break;
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
case LLM_ARCH_GROK:
{
result = llm.build_grok();
} break;
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
case LLM_ARCH_STARCODER:
{
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
result = llm.build_starcoder();
llama : add support for StarCoder model architectures (#3187) * add placeholder of starcoder in gguf / llama.cpp * support convert starcoder weights to gguf * convert MQA to MHA * fix ffn_down name * add LLM_ARCH_STARCODER to llama.cpp * set head_count_kv = 1 * load starcoder weight * add max_position_embeddings * set n_positions to max_positioin_embeddings * properly load all starcoder params * fix head count kv * fix comments * fix vram calculation for starcoder * store mqa directly * add input embeddings handling * add TBD * working in cpu, metal buggy * cleanup useless code * metal : fix out-of-bounds access in soft_max kernels * llama : make starcoder graph build more consistent with others * refactor: cleanup comments a bit * add other starcoder models: 3B, 7B, 15B * support-mqa-directly * fix: remove max_position_embeddings, use n_train_ctx * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * fix: switch to space from tab --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-15 19:02:13 +00:00
} break;
llm : support Adept Persimmon 8B (#3410) * Produces garbage output * wip: correct tensors up to RoPE * correct tensors thru RoPE * Correct outputs through masked & softmax'd KQ * fp32 works * Rename adept->persimmon * Produces correct outputs * clean up convert scripts * remove printing logic from ggml.c * remove prints from llama.cpp & fix merge * trivial cleanups * Add offload funcs * update conversion script to directly take adept artifacts rather than .saftensors file * Fix norm eps bug * Support sqr and concat on metal, persimmon-8b-q4 runs correctly * Small changes from review * Formatting changes * Minor changes to conversion script * Remove old script * Fix editorconfig formatting * Fix build * add overlooked offload code ggml-ci
2023-10-07 07:12:43 +00:00
case LLM_ARCH_PERSIMMON:
{
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
result = llm.build_persimmon();
llama : fix missing break in Persimmon arch case statements (#3535)
2023-10-08 05:22:17 +00:00
} break;
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
case LLM_ARCH_REFACT:
{
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
result = llm.build_refact();
llm : add Refact model (#3329) * add refact model * resolve comments * rebase to the latest * solve alibi cpu error --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-04 13:23:39 +00:00
} break;
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
case LLM_ARCH_BERT:
llama : add support for Nomic Embed (#5468)
2024-02-13 17:03:53 +00:00
case LLM_ARCH_NOMIC_BERT:
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
{
result = llm.build_bert();
} break;
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
case LLM_ARCH_BLOOM:
{
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
result = llm.build_bloom();
llm : add bloom models (#3553) * feat: Support bloom models * fix(bloom): fix model size --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 14:48:21 +00:00
} break;
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
case LLM_ARCH_MPT:
{
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
result = llm.build_mpt();
llm : add MPT support (#3417) * CUDA: added support for ggml_clamp (see also: https://github.com/ggerganov/ggml/issues/545) * mpt : added an implementation based (mostly) on falcon integration, modified with deltas from ggml/examples/mpt * mpt : protect against "clip_qkv": null in mpt-7b * mpt : quick fix to avoid "Strange model" warning when quantizing MPT models * mpt : addendum to changeset:84e30e8 - leave parameter clamp_kqv out from metadata rather than use 0.0 to indicate "no clamping" (more compliant with the current GGUF spec?) * mpt : standardized all tensor names to follow GGUF spec * mpt : addendum to changeset:1be89c40 - use "req" parameter of GGUF_GET_KEY macro instead of duplicate code * mpt : fixed comment s/gptneox/mpt/ * mpt : remove tabs, trailing whitespace * mpt : removed ne01 + n_past == ne00 assertion from alibi (cuda/f32) and rope_shift from build_mpt * mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 * comment out n_past instead of marking it unused * mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] * mpt : remove unused tokenizer_json in convert script * ggml : remove obsolete n_past assert in ggml_alibi * llama : print clam_kqv and max_alibi_bias hparams --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-10 07:50:23 +00:00
} break;
stablelm : StableLM support (#3586) * Add support for stablelm-3b-4e1t * Supports GPU offloading of (n-1) layers
2023-11-14 10:17:12 +00:00
case LLM_ARCH_STABLELM:
{
result = llm.build_stablelm();
} break;
llama : add Qwen support (#4281) * enable qwen to llama.cpp * llama : do not GPU split bias tensors --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-01 18:16:31 +00:00
case LLM_ARCH_QWEN:
{
result = llm.build_qwen();
} break;
llama : support upcoming Qwen2 (#5037)
2024-01-19 11:53:13 +00:00
case LLM_ARCH_QWEN2:
{
result = llm.build_qwen2();
} break;
llama : add qwen2moe (#6074) * support qwen2moe * fix-review * metal : support unary ops for nelements % 4 != 0 * metal : require contiguousness for float4 unary kernels * metal : require contiguousness for float4 unary kernels (cont) * fix-review * names : for brevity "SHARED_EXP" -> "SHEXP" * llama : reuse build_moe_ffn() * llama : add model type name --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-16 15:40:48 +00:00
case LLM_ARCH_QWEN2MOE:
{
result = llm.build_qwen2moe();
} break;
llama : add phi-2 + fix NeoX rope + ggml_mul_mat_set_prec (#4490) * phi2 implementation * fix breaking change * phi-2 : various fixes * phi-2 : use layer norm eps * py : whitespaces * llama : fix meta KV override bug * convert : phi don't add BOS token * convert : revert "added_tokens_decoder" change * phi-2 : scale Q instead of KQ for better precision * ggml : fix NeoX rope to rotate just first n_dims * cuda : less diff in the rope_neox kernel * ggml : add ggml_mul_mat_set_prec ggml-ci * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * Update ggml-cuda.cu Co-authored-by: slaren <slarengh@gmail.com> * cuda : ggml_cuda_op_mul_mat_cublas support F32 precision * cuda : remove oboslete comment --------- Co-authored-by: Ebey Abraham <ebeyabraham@microsoft.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-12-18 17:27:47 +00:00
case LLM_ARCH_PHI2:
{
result = llm.build_phi2();
} break;
llama : add phi3 support (#6852) * add explicit phi3 support * add explicit phi3 support * remove unused code * convert : add BOS token * llama : match EOT token <|end|> * llama : minor / style * llama : tabs -> spaces * convert : fix lint checks --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-24 07:00:37 +00:00
case LLM_ARCH_PHI3:
{
result = llm.build_phi3();
} break;
llama : add PLaMo model (#3557) * add plamo mock * add tensor loading * plamo convert * update norm * able to compile * fix norm_rms_eps hparam * runnable * use inp_pos * seems ok * update kqv code * remove develop code * update README * shuffle attn_q.weight and attn_output.weight for broadcasting * remove plamo_llm_build_kqv and use llm_build_kqv * fix style * update * llama : remove obsolete KQ_scale * plamo : fix tensor names for correct GPU offload --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-24 13:35:49 +00:00
case LLM_ARCH_PLAMO:
{
result = llm.build_plamo();
} break;
gpt2 : Add gpt2 architecture integration (#4555)
2023-12-28 14:03:57 +00:00
case LLM_ARCH_GPT2:
{
result = llm.build_gpt2();
} break;
llama : add CodeShell support (#5016) * llama: add codeshell support * llama.cpp: fix codeshell with NeoX rope Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-19 09:07:27 +00:00
case LLM_ARCH_CODESHELL:
{
result = llm.build_codeshell();
} break;
llama : add support for Orion-14B (#5118) * add support for Orion-14B(https://huggingface.co/OrionStarAI/Orion-14B-Chat) * flake8 support * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * Update llama.cpp * Update llama.cpp --------- Co-authored-by: lixiaopu <lixiaopu@cmcm.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-01-28 08:00:30 +00:00
case LLM_ARCH_ORION:
{
result = llm.build_orion();
} break;
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
case LLM_ARCH_INTERNLM2:
{
result = llm.build_internlm2();
} break;
llama : add MiniCPM support (#5346) * support minicpm arch. * fix tab/space typo. * convert minicpm model via convert-hf-gguf.py * try to make tokenizer work * fix bug for quantize minicpm * fix for flake8 lint * remove convert-minicpm.py * fix for editorconfig * correct minicpm model type (size) * constants expanded for minicpm * Minor change of the constant names for minicpm
2024-02-07 06:15:56 +00:00
case LLM_ARCH_MINICPM:
{
result = llm.build_minicpm();
} break;
llama : add `gemma` model (#5631) There are couple things in this architecture: 1. Shared input and output embedding parameters. 2. Key length and value length are not derived from `n_embd`. More information about the models can be found at https://ai.google.dev/gemma. GGUFs can be downloaded from https://huggingface.co/google.
2024-02-21 13:08:22 +00:00
case LLM_ARCH_GEMMA:
{
result = llm.build_gemma();
} break;
llama : add StarCoder2 support (#5795) * Add support for starcoder2 * handle rope type * skip rope freq and rotary embeddings from being serialized * resolve comments * Update llama.cpp * remove redundant changes * handle `rope-theta` * llama : change starcoder2 rope type * address comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-01 19:30:46 +00:00
case LLM_ARCH_STARCODER2:
{
result = llm.build_starcoder2();
} break;
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
case LLM_ARCH_MAMBA:
{
result = llm.build_mamba();
} break;
[Model] Add support for xverse (#6301) * Support xverse model convert to gguf format. * 1. Convert xverse models to gguf; 2. Add LLM_ARCH_XVERSE inference in llama.cpp; 3. Add xverse item in Supported models in README.md; * * gguf-py: remove redundant logs * llama: remove the init_mapping_prefetch custom parameter * llama.cpp: Include the changes from #6122 to exclude the unused outputs of the last layers. * - Fix format issues - Remove duplicate set kqv_out to llm_build_kv * Update llama.cpp --------- Co-authored-by: willhe <willhe@xverse.cn> Co-authored-by: willhe <hexin@xverse.cn>
2024-03-29 13:37:03 +00:00
case LLM_ARCH_XVERSE:
{
result = llm.build_xverse();
} break;
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
case LLM_ARCH_COMMAND_R:
{
result = llm.build_command_r();
} break;
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
case LLM_ARCH_DBRX:
{
result = llm.build_dbrx();
} break;
Implement the OLMo architecture (#6741) * implement olmo architecture * remove unused variable * remove unused moe branch * remove check for weight * remove superfluous moe, bias and rope tensors * clarified comment * fix clamp_kqv setting * remove obsolete parameter name filter
2024-04-19 09:35:54 +00:00
case LLM_ARCH_OLMO:
{
result = llm.build_olmo();
} break;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
default:
GGML_ASSERT(false);
build : enable more non-default compiler warnings (#3200)
2023-09-28 21:41:44 +00:00
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llm : add llm_build_context (#3881) * llm : add llm_build_context * llm : deduce norm eps based on type + explict max_alibi_bias, clamp_kqv * llm : restore the non-graph llm_build_ functional API ggml-ci * llm : cleanup + comments
2023-11-01 18:11:02 +00:00
llm.free();
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
return result;
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
static void llama_set_k_shift(llama_context & lctx) {
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
const int64_t kv_size = lctx.kv_self.size;
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
assert(ggml_backend_buffer_is_host(lctx.inp_K_shift->buffer));
int32_t * data = (int32_t *) lctx.inp_K_shift->data;
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
for (int i = 0; i < kv_size; ++i) {
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
data[i] = lctx.kv_self.cells[i].delta;
}
}
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
static void llama_set_s_copy(llama_context & lctx) {
const int64_t kv_size = lctx.kv_self.size;
assert(ggml_backend_buffer_is_host(lctx.inp_s_copy->buffer));
int32_t * data = (int32_t *) lctx.inp_s_copy->data;
for (int i = 0; i < kv_size; ++i) {
data[i] = lctx.kv_self.cells[i].src;
}
}
sync : ggml (#5452) * ggml-alloc : v3 (ggml/727) * ggml-alloc v3 ggml-ci * fix ci ggml-ci * whisper : check for backend buffer allocation failures * whisper : avoid leaks when initialization fails * cleanup ggml-ci * style fixes ggml-ci * sync : ggml * update llama.cpp, clip.cpp, export-lora.cpp * update finetune.cpp, train-text-from-scratch.cpp ggml-ci * ggml-backend : reduce alignment to 32 to match gguf and fix mmap --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-12 07:16:06 +00:00
static void llama_set_inputs(llama_context & lctx, const llama_batch & batch) {
//
// set input data
//
const auto & hparams = lctx.model.hparams;
const auto & cparams = lctx.cparams;
const auto & kv_self = lctx.kv_self;
if (batch.token) {
const int64_t n_tokens = batch.n_tokens;
ggml_backend_tensor_set(lctx.inp_tokens, batch.token, 0, n_tokens*ggml_element_size(lctx.inp_tokens));
}
if (batch.embd) {
const int64_t n_embd = hparams.n_embd;
const int64_t n_tokens = batch.n_tokens;
ggml_backend_tensor_set(lctx.inp_embd, batch.embd, 0, n_tokens*n_embd*ggml_element_size(lctx.inp_embd));
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
if (batch.pos && lctx.inp_pos) {
sync : ggml (#5452) * ggml-alloc : v3 (ggml/727) * ggml-alloc v3 ggml-ci * fix ci ggml-ci * whisper : check for backend buffer allocation failures * whisper : avoid leaks when initialization fails * cleanup ggml-ci * style fixes ggml-ci * sync : ggml * update llama.cpp, clip.cpp, export-lora.cpp * update finetune.cpp, train-text-from-scratch.cpp ggml-ci * ggml-backend : reduce alignment to 32 to match gguf and fix mmap --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-12 07:16:06 +00:00
const int64_t n_tokens = batch.n_tokens;
ggml_backend_tensor_set(lctx.inp_pos, batch.pos, 0, n_tokens*ggml_element_size(lctx.inp_pos));
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (hparams.causal_attn || cparams.pooling_type == LLAMA_POOLING_TYPE_NONE) {
GGML_ASSERT(lctx.inp_out_ids && "every model that can must skip unused outputs");
const int64_t n_tokens = batch.n_tokens;
GGML_ASSERT(ggml_backend_buffer_is_host(lctx.inp_out_ids->buffer));
int32_t * data = (int32_t *) lctx.inp_out_ids->data;
if (lctx.n_outputs == n_tokens) {
for (int i = 0; i < n_tokens; ++i) {
data[i] = i;
}
} else if (batch.logits) {
int32_t n_outputs = 0;
for (int i = 0; i < n_tokens; ++i) {
if (batch.logits[i]) {
data[n_outputs++] = i;
}
}
// the graph needs to have been passed the correct number of outputs
GGML_ASSERT(lctx.n_outputs == n_outputs);
} else if (lctx.n_outputs == 1) {
// only keep last output
data[0] = n_tokens - 1;
} else {
GGML_ASSERT(lctx.n_outputs == 0);
}
}
llama : add support for GritLM (#5959) * add gritlm example * gritlm results match * tabs to spaces * comment out debug printing * rebase to new embed * gritlm embeddings are back babeee * add to gitignore * allow to toggle embedding mode * Clean-up GritLM sample code. * Fix types. * Flush stdout and output ending newline if streaming. * mostly style fixes; correct KQ_mask comment * add causal_attn flag to llama_cparams * gritml : minor * llama : minor --------- Co-authored-by: Douglas Hanley <thesecretaryofwar@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-10 15:56:30 +00:00
GGML_ASSERT(
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
// (!a || b) is a logical implication (a -> b)
// !hparams.causal_attn -> !cparams.causal_attn
llama : add support for GritLM (#5959) * add gritlm example * gritlm results match * tabs to spaces * comment out debug printing * rebase to new embed * gritlm embeddings are back babeee * add to gitignore * allow to toggle embedding mode * Clean-up GritLM sample code. * Fix types. * Flush stdout and output ending newline if streaming. * mostly style fixes; correct KQ_mask comment * add causal_attn flag to llama_cparams * gritml : minor * llama : minor --------- Co-authored-by: Douglas Hanley <thesecretaryofwar@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-10 15:56:30 +00:00
(hparams.causal_attn || !cparams.causal_attn) &&
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
"causal attention with embedding models is not supported"
llama : add support for GritLM (#5959) * add gritlm example * gritlm results match * tabs to spaces * comment out debug printing * rebase to new embed * gritlm embeddings are back babeee * add to gitignore * allow to toggle embedding mode * Clean-up GritLM sample code. * Fix types. * Flush stdout and output ending newline if streaming. * mostly style fixes; correct KQ_mask comment * add causal_attn flag to llama_cparams * gritml : minor * llama : minor --------- Co-authored-by: Douglas Hanley <thesecretaryofwar@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-10 15:56:30 +00:00
);
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
if (lctx.inp_KQ_mask) {
// NOTE: hparams.causal_attn indicates the model is capable of generation and uses the kv cache.
if (cparams.causal_attn) {
const int64_t n_kv = kv_self.n;
const int64_t n_tokens = batch.n_tokens;
sync : ggml (#5452) * ggml-alloc : v3 (ggml/727) * ggml-alloc v3 ggml-ci * fix ci ggml-ci * whisper : check for backend buffer allocation failures * whisper : avoid leaks when initialization fails * cleanup ggml-ci * style fixes ggml-ci * sync : ggml * update llama.cpp, clip.cpp, export-lora.cpp * update finetune.cpp, train-text-from-scratch.cpp ggml-ci * ggml-backend : reduce alignment to 32 to match gguf and fix mmap --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-12 07:16:06 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
GGML_ASSERT(ggml_backend_buffer_is_host(lctx.inp_KQ_mask->buffer));
sync : ggml (#5452) * ggml-alloc : v3 (ggml/727) * ggml-alloc v3 ggml-ci * fix ci ggml-ci * whisper : check for backend buffer allocation failures * whisper : avoid leaks when initialization fails * cleanup ggml-ci * style fixes ggml-ci * sync : ggml * update llama.cpp, clip.cpp, export-lora.cpp * update finetune.cpp, train-text-from-scratch.cpp ggml-ci * ggml-backend : reduce alignment to 32 to match gguf and fix mmap --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-12 07:16:06 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
float * data = (float *) lctx.inp_KQ_mask->data;
sync : ggml (#5452) * ggml-alloc : v3 (ggml/727) * ggml-alloc v3 ggml-ci * fix ci ggml-ci * whisper : check for backend buffer allocation failures * whisper : avoid leaks when initialization fails * cleanup ggml-ci * style fixes ggml-ci * sync : ggml * update llama.cpp, clip.cpp, export-lora.cpp * update finetune.cpp, train-text-from-scratch.cpp ggml-ci * ggml-backend : reduce alignment to 32 to match gguf and fix mmap --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-12 07:16:06 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// For causal attention, use only the previous KV cells
// of the correct sequence for each token of the batch.
// It's assumed that if a token in the batch has multiple sequences, they are equivalent.
for (int h = 0; h < 1; ++h) {
for (int j = 0; j < n_tokens; ++j) {
const llama_pos pos = batch.pos[j];
const llama_seq_id seq_id = batch.seq_id[j][0];
sync : ggml (#5452) * ggml-alloc : v3 (ggml/727) * ggml-alloc v3 ggml-ci * fix ci ggml-ci * whisper : check for backend buffer allocation failures * whisper : avoid leaks when initialization fails * cleanup ggml-ci * style fixes ggml-ci * sync : ggml * update llama.cpp, clip.cpp, export-lora.cpp * update finetune.cpp, train-text-from-scratch.cpp ggml-ci * ggml-backend : reduce alignment to 32 to match gguf and fix mmap --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-12 07:16:06 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
for (int i = 0; i < n_kv; ++i) {
float f;
if (!lctx.kv_self.cells[i].has_seq_id(seq_id) || lctx.kv_self.cells[i].pos > pos) {
f = -INFINITY;
} else {
f = 0.0f;
}
data[h*(n_kv*n_tokens) + j*n_kv + i] = f;
sync : ggml (#5452) * ggml-alloc : v3 (ggml/727) * ggml-alloc v3 ggml-ci * fix ci ggml-ci * whisper : check for backend buffer allocation failures * whisper : avoid leaks when initialization fails * cleanup ggml-ci * style fixes ggml-ci * sync : ggml * update llama.cpp, clip.cpp, export-lora.cpp * update finetune.cpp, train-text-from-scratch.cpp ggml-ci * ggml-backend : reduce alignment to 32 to match gguf and fix mmap --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-12 07:16:06 +00:00
}
}
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
} else {
// when using kv cache, the mask needs to match the kv cache size
const int64_t n_tokens = batch.n_tokens;
const int64_t n_stride = hparams.causal_attn ? kv_self.n : n_tokens;
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
GGML_ASSERT(ggml_backend_buffer_is_host(lctx.inp_KQ_mask->buffer));
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
float * data = (float *) lctx.inp_KQ_mask->data;
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
for (int h = 0; h < 1; ++h) {
for (int j = 0; j < n_tokens; ++j) {
const llama_seq_id seq_id = batch.seq_id[j][0];
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
for (int i = 0; i < n_tokens; ++i) {
float f = -INFINITY;
for (int s = 0; s < batch.n_seq_id[i]; ++s) {
if (batch.seq_id[i][s] == seq_id) {
f = 0.0f;
break;
}
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
data[h*(n_tokens*n_tokens) + j*n_stride + i] = f;
}
llama : add support for GritLM (#5959) * add gritlm example * gritlm results match * tabs to spaces * comment out debug printing * rebase to new embed * gritlm embeddings are back babeee * add to gitignore * allow to toggle embedding mode * Clean-up GritLM sample code. * Fix types. * Flush stdout and output ending newline if streaming. * mostly style fixes; correct KQ_mask comment * add causal_attn flag to llama_cparams * gritml : minor * llama : minor --------- Co-authored-by: Douglas Hanley <thesecretaryofwar@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-10 15:56:30 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
for (int i = n_tokens; i < n_stride; ++i) {
data[h*(n_tokens*n_tokens) + j*n_stride + i] = -INFINITY;
}
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
}
}
}
sync : ggml (#5452) * ggml-alloc : v3 (ggml/727) * ggml-alloc v3 ggml-ci * fix ci ggml-ci * whisper : check for backend buffer allocation failures * whisper : avoid leaks when initialization fails * cleanup ggml-ci * style fixes ggml-ci * sync : ggml * update llama.cpp, clip.cpp, export-lora.cpp * update finetune.cpp, train-text-from-scratch.cpp ggml-ci * ggml-backend : reduce alignment to 32 to match gguf and fix mmap --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-12 07:16:06 +00:00
}
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
// ALiBi requires the KQ_pos tensor to provide the sequence position of each token in the batch
// this allows to process multiple sequences in parallel with ALiBi-based models
if (hparams.use_alibi) {
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
const int64_t n_kv = kv_self.n;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
GGML_ASSERT(lctx.inp_KQ_pos);
GGML_ASSERT(ggml_backend_buffer_is_host(lctx.inp_KQ_pos->buffer));
ggml : add ALiBi support for ggml_soft_max_ext (#5488) * ggml : avoid recomputing alibi slopes (CPU) * llama : reuse hparams.f_max_alibi_bias in all cases ggml-ci * ggml : support alibi bias in ggml_soft_max_ext (CPU + Metal) ggml-ci * ggml : handle all SRCs (do not break on first null) ggml-ci * tests : do not use slope for large soft_max accumulates too much error ggml-ci * ggml : alternative ALiBi without extra tensor We compute the slopes in the kernel ggml-ci * cuda : add ALiBi support in ggml_soft_max_ext ggml-ci * ggml : deprecate ggml_alibi * ggml : support multi-sequence ALiBi (Metal) ggml-ci * cuda : add multi-seq ALiBi + remote F16 soft_max ggml-ci * ggml : update deprecation message * ggml : fix pos ptr when no ALiBi ggml-ci * cuda : fix performance (pow -> powf) * cuda : precompute ALiBi constants * metal : pre-compute ALiBi slopes ggml-ci * llama : init kq_pos only if needed ggml-ci * test-backend-ops : add null pos test to soft_max test-backend-ops : replace soft_max tests ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-17 21:04:16 +00:00
float * data = (float *) lctx.inp_KQ_pos->data;
for (int i = 0; i < n_kv; ++i) {
data[i] = float(lctx.kv_self.cells[i].pos);
}
}
llama : allow for user specified embedding pooling type (#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-03 10:40:27 +00:00
if (cparams.pooling_type == LLAMA_POOLING_TYPE_MEAN) {
llama : support batched embeddings (#5466) * batched embedding: pool outputs by sequence id. updated embedding example * bring back non-causal attention * embd : minor improvements * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-13 12:06:58 +00:00
const int64_t n_tokens = batch.n_tokens;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
GGML_ASSERT(lctx.inp_mean);
Use correct type of pooling for embedding models (#5500) Use correct type of pooling for embedding models
2024-02-15 17:21:49 +00:00
GGML_ASSERT(ggml_backend_buffer_is_host(lctx.inp_mean->buffer));
llama : support batched embeddings (#5466) * batched embedding: pool outputs by sequence id. updated embedding example * bring back non-causal attention * embd : minor improvements * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-13 12:06:58 +00:00
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
float * data = (float *) lctx.inp_mean->data;
Use correct type of pooling for embedding models (#5500) Use correct type of pooling for embedding models
2024-02-15 17:21:49 +00:00
memset(lctx.inp_mean->data, 0, n_tokens * n_tokens * ggml_element_size(lctx.inp_mean));
llama : support batched embeddings (#5466) * batched embedding: pool outputs by sequence id. updated embedding example * bring back non-causal attention * embd : minor improvements * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-13 12:06:58 +00:00
Use correct type of pooling for embedding models (#5500) Use correct type of pooling for embedding models
2024-02-15 17:21:49 +00:00
std::vector<uint64_t> sum(n_tokens, 0);
llama : support batched embeddings (#5466) * batched embedding: pool outputs by sequence id. updated embedding example * bring back non-causal attention * embd : minor improvements * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-13 12:06:58 +00:00
for (int i = 0; i < n_tokens; ++i) {
const llama_seq_id seq_id = batch.seq_id[i][0];
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
GGML_ASSERT(seq_id < n_tokens && "seq_id cannot be larger than n_tokens with pooling_type == MEAN");
Use correct type of pooling for embedding models (#5500) Use correct type of pooling for embedding models
2024-02-15 17:21:49 +00:00
sum[seq_id] += 1;
}
std::vector<float> div(n_tokens, 0.0f);
for (int i = 0; i < n_tokens; ++i) {
const uint64_t s = sum[i];
if (s > 0) {
div[i] = 1.0f/float(s);
}
}
for (int i = 0; i < n_tokens; ++i) {
const llama_seq_id seq_id = batch.seq_id[i][0];
data[seq_id*n_tokens + i] = div[seq_id];
}
}
llama : allow for user specified embedding pooling type (#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-03 10:40:27 +00:00
if (cparams.pooling_type == LLAMA_POOLING_TYPE_CLS) {
Use correct type of pooling for embedding models (#5500) Use correct type of pooling for embedding models
2024-02-15 17:21:49 +00:00
const int64_t n_tokens = batch.n_tokens;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
GGML_ASSERT(lctx.inp_cls);
Use correct type of pooling for embedding models (#5500) Use correct type of pooling for embedding models
2024-02-15 17:21:49 +00:00
GGML_ASSERT(ggml_backend_buffer_is_host(lctx.inp_cls->buffer));
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
Use correct type of pooling for embedding models (#5500) Use correct type of pooling for embedding models
2024-02-15 17:21:49 +00:00
uint32_t * data = (uint32_t *) lctx.inp_cls->data;
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
memset(lctx.inp_cls->data, 0, n_tokens * ggml_element_size(lctx.inp_cls));
Use correct type of pooling for embedding models (#5500) Use correct type of pooling for embedding models
2024-02-15 17:21:49 +00:00
for (int i = 0; i < n_tokens; ++i) {
const llama_seq_id seq_id = batch.seq_id[i][0];
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
const llama_pos pos = batch.pos[i];
GGML_ASSERT(seq_id < n_tokens && "seq_id cannot be larger than n_tokens with pooling_type == CLS");
Use correct type of pooling for embedding models (#5500) Use correct type of pooling for embedding models
2024-02-15 17:21:49 +00:00
if (pos == 0) {
data[seq_id] = i;
}
llama : support batched embeddings (#5466) * batched embedding: pool outputs by sequence id. updated embedding example * bring back non-causal attention * embd : minor improvements * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-13 12:06:58 +00:00
}
}
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
if (kv_self.recurrent) {
const int64_t n_kv = kv_self.n;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
if (lctx.inp_s_mask) {
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
GGML_ASSERT(ggml_backend_buffer_is_host(lctx.inp_s_mask->buffer));
float * data = (float *) lctx.inp_s_mask->data;
// states which are not affected by the current batch are left untouched
for (int i = 0; i < n_kv; ++i) {
llama_seq_id seq_id = i + lctx.kv_self.head;
llama_kv_cell & kv_cell = lctx.kv_self.cells[seq_id];
bool has_self_seq = kv_cell.has_seq_id(seq_id);
data[i] = (float) has_self_seq;
// ensure current sequences will be kept
if (!has_self_seq && kv_cell.pos >= 0) {
kv_cell.seq_id.insert(seq_id);
}
}
}
// For Mamba (and other recurrent architectures),
// update the correct state(s)/sequence(s) for each token of the batch.
// Like with the KQ_mask, if a token in the batch has multiple sequences,
// they are assumed to be equivalent (not here, but in ggml_ssm_scan and ggml_ssm_conv).
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
if (lctx.inp_s_seq) {
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
const int64_t n_tokens = batch.n_tokens;
GGML_ASSERT(ggml_backend_buffer_is_host(lctx.inp_s_seq->buffer));
int32_t * data = (int32_t *) lctx.inp_s_seq->data;
for (int j = 0; j < n_tokens; ++j) {
const int32_t n_seq = batch.n_seq_id[j];
GGML_ASSERT(0 < n_seq); // a token should be part of at least 1 sequence
for (int i = 0; i < n_kv; ++i) {
if (i < n_seq) {
// for this type of model, the head is the minimum seq_id of the batch
data[j*n_kv + i] = batch.seq_id[j][i] - kv_self.head;
} else {
data[j*n_kv + i] = -1;
}
}
}
}
}
sync : ggml (#5452) * ggml-alloc : v3 (ggml/727) * ggml-alloc v3 ggml-ci * fix ci ggml-ci * whisper : check for backend buffer allocation failures * whisper : avoid leaks when initialization fails * cleanup ggml-ci * style fixes ggml-ci * sync : ggml * update llama.cpp, clip.cpp, export-lora.cpp * update finetune.cpp, train-text-from-scratch.cpp ggml-ci * ggml-backend : reduce alignment to 32 to match gguf and fix mmap --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-12 07:16:06 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
// Make sure enough space is available for outputs.
// Returns max number of outputs for which space was reserved.
static size_t llama_output_reserve(llama_context & lctx, size_t n_outputs) {
const auto & cparams = lctx.cparams;
const auto & hparams = lctx.model.hparams;
const size_t n_outputs_max = std::max(n_outputs, (size_t) cparams.n_seq_max);
const auto n_batch = cparams.n_batch;
const auto n_vocab = hparams.n_vocab;
const auto n_embd = hparams.n_embd;
// TODO: use a per-batch flag for logits presence instead
const bool has_logits = cparams.causal_attn;
const bool has_embd = cparams.embeddings && (hparams.causal_attn || cparams.pooling_type == LLAMA_POOLING_TYPE_NONE);
const size_t logits_size = has_logits ? n_vocab*n_outputs_max : 0;
const size_t embd_size = has_embd ? n_embd*n_outputs_max : 0;
if (lctx.output_ids.empty()) {
// init, never resized afterwards
lctx.output_ids.resize(n_batch);
}
const size_t prev_size = lctx.buf_output ? ggml_backend_buffer_get_size(lctx.buf_output) : 0;
const size_t new_size = (logits_size + embd_size) * sizeof(float);
// alloc only when more than the current capacity is required
// TODO: also consider shrinking the buffer
if (!lctx.buf_output || prev_size < new_size) {
if (lctx.buf_output) {
#ifndef NDEBUG
// This doesn't happen often, but may be annoying in some cases (like the HellaSwag benchmark)
LLAMA_LOG_INFO("%s: reallocating output buffer from size %.02f MiB to %.02f MiB\n", __func__, prev_size / 1024.0 / 1024.0, new_size / 1024.0 / 1024.0);
#endif
ggml_backend_buffer_free(lctx.buf_output);
lctx.buf_output = nullptr;
lctx.logits = nullptr;
lctx.embd = nullptr;
}
lctx.buf_output = ggml_backend_buft_alloc_buffer(llama_default_buffer_type_cpu(true), new_size);
if (lctx.buf_output == nullptr) {
LLAMA_LOG_ERROR("%s: failed to allocate output buffer of size %.2f MiB\n", __func__, new_size / (1024.0 * 1024.0));
return 0;
}
}
float * output_base = (float *) ggml_backend_buffer_get_base(lctx.buf_output);
lctx.logits = has_logits ? output_base : nullptr;
lctx.embd = has_embd ? output_base + logits_size : nullptr;
lctx.output_size = n_outputs_max;
lctx.logits_size = logits_size;
lctx.embd_size = embd_size;
// set all ids as invalid (negative)
std::fill(lctx.output_ids.begin(), lctx.output_ids.end(), -1);
ggml_backend_buffer_clear(lctx.buf_output, 0);
lctx.n_outputs = 0;
return n_outputs_max;
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
static void llama_graph_compute(
llama_context & lctx,
ggml_cgraph * gf,
int n_threads) {
#ifdef GGML_USE_MPI
const int64_t n_layer = lctx.model.hparams.n_layer;
ggml_mpi_graph_compute_pre(lctx.ctx_mpi, gf, n_layer);
#endif
#ifdef GGML_USE_METAL
if (ggml_backend_is_metal(lctx.backend_metal)) {
ggml_backend_metal_set_n_cb(lctx.backend_metal, n_threads);
}
#endif
if (lctx.backend_cpu != nullptr) {
ggml_backend_cpu_set_n_threads(lctx.backend_cpu, n_threads);
llama : add abort_callback to interrupt computation (#5409) * using abort_callback from ggml to stop llama computation * format fix * a brief explaining comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-02 19:52:25 +00:00
ggml_backend_cpu_set_abort_callback(lctx.backend_cpu, lctx.abort_callback, lctx.abort_callback_data);
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
ggml_backend_sched_graph_compute_async(lctx.sched, gf);
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
// fprintf(stderr, "splits: %d\n", ggml_backend_sched_get_n_splits(lctx.sched));
#ifdef GGML_USE_MPI
ggml_mpi_graph_compute_post(lctx.ctx_mpi, gf, n_layer);
#endif
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
// decode a batch of tokens by evaluating the transformer
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
//
// - lctx: llama context
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
// - batch: batch to evaluate
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
//
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
// return 0 on success
// return positive int on warning
// return negative int on error
//
static int llama_decode_internal(
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
llama_context & lctx,
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
llama_batch batch_all) { // TODO: rename back to batch
const uint32_t n_tokens_all = batch_all.n_tokens;
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
if (n_tokens_all == 0) {
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
LLAMA_LOG_ERROR("%s: n_tokens == 0", __func__);
return -1;
}
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
const auto & model = lctx.model;
const auto & hparams = model.hparams;
const auto & cparams = lctx.cparams;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
GGML_ASSERT((!batch_all.token && batch_all.embd) || (batch_all.token && !batch_all.embd)); // NOLINT
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
GGML_ASSERT(n_tokens_all <= cparams.n_batch);
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
GGML_ASSERT((cparams.causal_attn || cparams.n_ubatch >= n_tokens_all) && "non-causal attention requires n_ubatch >= n_tokens");
llama : add benchmark example (#2626) * llama : add benchmark example * add to examples CMakeLists.txt * fix msvc build * add missing include * add Bessel's correction to stdev calculation Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * improve markdown formatting * add missing include * print warning is NDEBUG is not defined * remove n_prompt and n_gen from the matrix, use each value separately instead * better checks for non-optimized builds * llama.cpp : fix MEM_REQ_SCRATCH0 reusing the value of n_ctx of the first call * fix json formatting * add sql output * add basic cpu and gpu info (linx/cuda only) * markdown: also show values that differ from the default * markdown: add build id * cleanup * improve formatting * formatting --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2023-08-18 10:44:58 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
if (lctx.t_compute_start_us == 0) {
lctx.t_compute_start_us = ggml_time_us();
}
lctx.n_queued_tokens += n_tokens_all;
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
#ifdef GGML_USE_MPI
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
// TODO: needs fix after #3228
GGML_ASSERT(false && "not implemented");
//ggml_mpi_eval_init(lctx.ctx_mpi, &n_tokens, &n_past, &n_threads);
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
#endif
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
auto & kv_self = lctx.kv_self;
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
const int64_t n_embd = hparams.n_embd;
const int64_t n_vocab = hparams.n_vocab;
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
uint32_t n_outputs = 0;
uint32_t n_outputs_prev = 0;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
const auto n_ubatch = cparams.n_ubatch;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
std::vector<llama_pos> pos;
std::vector<int32_t> n_seq_id;
std::vector<llama_seq_id *> seq_id_arr;
std::vector<std::vector<llama_seq_id>> seq_id;
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
// count outputs
if (batch_all.logits) {
for (uint32_t i = 0; i < n_tokens_all; ++i) {
n_outputs += batch_all.logits[i] != 0;
}
} else if (lctx.logits_all || (cparams.embeddings && cparams.pooling_type != LLAMA_POOLING_TYPE_NONE)) {
n_outputs = n_tokens_all;
} else {
// keep last output only
n_outputs = 1;
}
// reserve output buffer
if (llama_output_reserve(lctx, n_outputs) < n_outputs) {
LLAMA_LOG_ERROR("%s: could not reserve space for batch with %u outputs\n", __func__, n_outputs);
return -2;
};
// set output mappings
if (batch_all.logits) {
int32_t i_logits = 0;
for (uint32_t i = 0; i < n_tokens_all; ++i) {
if (batch_all.logits[i]) {
lctx.output_ids[i] = i_logits++;
}
}
} else {
for (uint32_t i = 0; i < n_outputs; ++i) {
lctx.output_ids[i] = i;
}
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
for (uint32_t cur_token = 0; cur_token < n_tokens_all; cur_token += n_ubatch) {
const uint32_t n_tokens = std::min(n_ubatch, n_tokens_all - cur_token);
llama_batch u_batch = {
/* .n_tokens = */ (int32_t) n_tokens,
/* .token = */ batch_all.token ? batch_all.token + cur_token : nullptr,
/* .embd = */ batch_all.embd ? batch_all.embd + cur_token*n_embd : nullptr,
/* .pos = */ batch_all.pos ? batch_all.pos + cur_token : nullptr,
/* .n_seq_id = */ batch_all.n_seq_id ? batch_all.n_seq_id + cur_token : nullptr,
/* .seq_id = */ batch_all.seq_id ? batch_all.seq_id + cur_token : nullptr,
/* .logits = */ batch_all.logits ? batch_all.logits + cur_token : nullptr,
/* .all_pos_0 = */ batch_all.all_pos_0 + (llama_pos) cur_token*batch_all.all_pos_1,
/* .all_pos_1 = */ batch_all.all_pos_1,
/* .all_seq_id = */ batch_all.all_seq_id,
};
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
// count the outputs in this u_batch
{
int32_t n_outputs_new = 0;
if (u_batch.logits) {
for (uint32_t i = 0; i < n_tokens; i++) {
n_outputs_new += u_batch.logits[i] != 0;
}
} else if (n_outputs == n_tokens_all) {
n_outputs_new = n_tokens;
} else {
// keep last output only
if (cur_token + n_tokens >= n_tokens_all) {
n_outputs_new = 1;
}
}
// needs to happen before the graph is built
lctx.n_outputs = n_outputs_new;
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
int n_threads = n_tokens == 1 ? cparams.n_threads : cparams.n_threads_batch;
GGML_ASSERT(n_threads > 0);
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// helpers for smoother batch API transition
// after deprecating the llama_eval calls, these will be removed
if (u_batch.pos == nullptr) {
pos.resize(n_tokens);
for (uint32_t i = 0; i < n_tokens; i++) {
pos[i] = u_batch.all_pos_0 + i*u_batch.all_pos_1;
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
u_batch.pos = pos.data();
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
if (u_batch.seq_id == nullptr) {
n_seq_id.resize(n_tokens);
seq_id.resize(n_tokens);
seq_id_arr.resize(n_tokens);
for (uint32_t i = 0; i < n_tokens; i++) {
n_seq_id[i] = 1;
seq_id[i].resize(1);
seq_id[i][0] = u_batch.all_seq_id;
seq_id_arr[i] = seq_id[i].data();
}
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
u_batch.n_seq_id = n_seq_id.data();
u_batch.seq_id = seq_id_arr.data();
}
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// non-causal masks do not use the KV cache
if (hparams.causal_attn) {
llama_kv_cache_update(&lctx);
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// if we have enough unused cells before the current head ->
// better to start searching from the beginning of the cache, hoping to fill it
if (kv_self.head > kv_self.used + 2*n_tokens) {
kv_self.head = 0;
}
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
if (!llama_kv_cache_find_slot(kv_self, u_batch)) {
return 1;
}
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
if (!kv_self.recurrent) {
// a heuristic, to avoid attending the full cache if it is not yet utilized
// after enough generations, the benefit from this heuristic disappears
// if we start defragmenting the cache, the benefit from this will be more important
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
kv_self.n = std::min(kv_self.size, std::max(256u, GGML_PAD(llama_kv_cache_cell_max(kv_self), 256)));
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
//kv_self.n = llama_kv_cache_cell_max(kv_self);
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
}
}
fix embeddings when using CUDA (#3657)
2023-10-17 20:24:50 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
//printf("kv_self.n = %5d, kv_self.used = %5d, kv_self.head = %5d\n", kv_self.n, kv_self.used, kv_self.head);
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
ggml_backend_sched_reset(lctx.sched);
ggml_backend_sched_set_eval_callback(lctx.sched, lctx.cparams.cb_eval, lctx.cparams.cb_eval_user_data);
ggml : improve graph build time via hash table lookup (#2329) * improve graph build time * ggml_tensor : use 1 bit per flag * use a hash table instead
2023-07-25 12:32:20 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
ggml_cgraph * gf = llama_build_graph(lctx, u_batch, false);
sync : ggml (#5452) * ggml-alloc : v3 (ggml/727) * ggml-alloc v3 ggml-ci * fix ci ggml-ci * whisper : check for backend buffer allocation failures * whisper : avoid leaks when initialization fails * cleanup ggml-ci * style fixes ggml-ci * sync : ggml * update llama.cpp, clip.cpp, export-lora.cpp * update finetune.cpp, train-text-from-scratch.cpp ggml-ci * ggml-backend : reduce alignment to 32 to match gguf and fix mmap --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-12 07:16:06 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// the output is always the last tensor in the graph
struct ggml_tensor * res = gf->nodes[gf->n_nodes - 1];
struct ggml_tensor * embd = gf->nodes[gf->n_nodes - 2];
mpi : add support for distributed inference via MPI (#2099) * MPI support, first cut * fix warnings, update README * fixes * wrap includes * PR comments * Update CMakeLists.txt * Add GH workflow, fix test * Add info to README * mpi : trying to move more MPI stuff into ggml-mpi (WIP) (#2099) * mpi : add names for layer inputs + prep ggml_mpi_graph_compute() * mpi : move all MPI logic into ggml-mpi Not tested yet * mpi : various fixes - communication now works but results are wrong * mpi : fix output tensor after MPI compute (still not working) * mpi : fix inference * mpi : minor * Add OpenMPI to GH action * [mpi] continue-on-error: true * mpi : fix after master merge * [mpi] Link MPI C++ libraries to fix OpenMPI * tests : fix new llama_backend API * [mpi] use MPI_INT32_T * mpi : factor out recv / send in functions and reuse * mpi : extend API to allow usage with outer backends (e.g. Metal) --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-10 15:49:56 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (lctx.n_outputs == 0) {
// no output
res = nullptr;
embd = nullptr;
} else if (!hparams.causal_attn) {
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
res = nullptr; // do not extract logits for embedding models such as BERT
llama : fix kv shift bug (#3835) ggml-ci
2023-10-29 16:32:51 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// token or sequence embeddings
embd = gf->nodes[gf->n_nodes - 1];
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
GGML_ASSERT(strcmp(embd->name, "result_embd") == 0 || strcmp(embd->name, "result_embd_pooled") == 0);
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
} else if (cparams.embeddings) {
// the embeddings could be in the second to last tensor, or any of the previous tensors
int i_embd = gf->n_nodes - 2;
for (int i = 3; strcmp(embd->name, "result_norm") != 0; ++i) {
i_embd = gf->n_nodes - i;
if (i_embd < 0) { break; }
embd = gf->nodes[i_embd];
}
GGML_ASSERT(i_embd >= 0 && "missing result_norm tensor");
// TODO: use a per-batch flag to know when to skip logits while keeping embeddings
if (!cparams.causal_attn) {
res = nullptr; // do not extract logits when not needed
// skip computing logits
// TODO: is this safe?
gf->n_nodes = i_embd + 1;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
} else {
embd = nullptr; // do not extract embeddings when not needed
GGML_ASSERT(strcmp(res->name, "result_output") == 0 && "missing result_output tensor");
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
}
// LLAMA_LOG_INFO("graph build time: %.3f ms (%d nodes, %d leafs)\n", (ggml_time_us() - t_start_us)/1000.0, gf->n_nodes, gf->n_leafs);
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// for big prompts, if BLAS is enabled, it is better to use only one thread
// otherwise, the threads are spin-lock waiting for the BLAS calls and are degrading the performance
// TODO: this is mostly important for Apple Silicon where CBLAS is still performing very well
// we still need some threads to process all non-mul_mat ops, but not too much to avoid interfering
// with the BLAS calls. need a better solution
// MoE Special Case: This logic applies when hparams.n_expert == 0, i.e. the model is NOT an MoE model. When an MoE is
// being processed then Accelerate/BLAS will not be involved, so capping would limit performance.
if (n_tokens >= 32 && hparams.n_expert == 0 && ggml_cpu_has_blas() && !ggml_cpu_has_gpublas()) {
n_threads = std::min(4, n_threads);
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
ggml_backend_sched_alloc_graph(lctx.sched, gf);
ggml, llama : avoid heavy V transpose + improvements (#775) ggml : - added ggml_view_3d() - ggml_view_tensor() now inherits the stride too - reimplement ggml_cpy() to account for dst stride - no longer require tensor->data to be memory aligned llama : - compute RoPE on 32-bit tensors (should be more accurate) - store RoPE-ed K in the KV cache - store transposed V in the KV cache (significant speed-up) - avoid unnecessary Q copy
2023-04-05 19:07:33 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
llama_set_inputs(lctx, u_batch);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
llama_graph_compute(lctx, gf, n_threads);
Add embedding mode with arg flag. Currently working (#282) * working but ugly * add arg flag, not working on embedding mode * typo * Working! Thanks to @nullhook * make params argument instead of hardcoded boolean. remove useless time check * start doing the instructions but not finished. This probably doesnt compile * Embeddings extraction support --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-03-24 15:05:13 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// update the kv ring buffer
{
kv_self.head += n_tokens;
llama : sanity checks for access to logits (#4274) Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-16 03:16:15 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// Ensure kv cache head points to a valid index.
if (kv_self.head >= kv_self.size) {
kv_self.head = 0;
}
}
llama : sanity checks for access to logits (#4274) Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-16 03:16:15 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
#ifdef GGML_PERF
// print timing information per ggml operation (for debugging purposes)
// requires GGML_PERF to be defined
ggml_graph_print(gf);
#endif
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// plot the computation graph in dot format (for debugging purposes)
//if (n_past%100 == 0) {
// ggml_graph_dump_dot(gf, NULL, "llama.dot");
//}
// extract logits
if (res) {
ggml_backend_t backend_res = ggml_backend_sched_get_tensor_backend(lctx.sched, res);
GGML_ASSERT(backend_res != nullptr);
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
GGML_ASSERT(lctx.logits != nullptr);
float * logits_out = lctx.logits + n_outputs_prev*n_vocab;
const int32_t n_outputs_new = lctx.n_outputs;
if (n_outputs_new) {
GGML_ASSERT( n_outputs_prev + n_outputs_new <= n_outputs);
GGML_ASSERT((n_outputs_prev + n_outputs_new)*n_vocab <= (int64_t) lctx.logits_size);
ggml_backend_tensor_get_async(backend_res, res, logits_out, 0, n_outputs_new*n_vocab*sizeof(float));
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
}
Add embedding mode with arg flag. Currently working (#282) * working but ugly * add arg flag, not working on embedding mode * typo * Working! Thanks to @nullhook * make params argument instead of hardcoded boolean. remove useless time check * start doing the instructions but not finished. This probably doesnt compile * Embeddings extraction support --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-03-24 15:05:13 +00:00
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// extract embeddings
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (embd) {
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
ggml_backend_t backend_embd = ggml_backend_sched_get_tensor_backend(lctx.sched, embd);
GGML_ASSERT(backend_embd != nullptr);
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
switch (cparams.pooling_type) {
case LLAMA_POOLING_TYPE_NONE:
{
// extract token embeddings
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
GGML_ASSERT(lctx.embd != nullptr);
float * embd_out = lctx.embd + n_outputs_prev*n_embd;
const int32_t n_outputs_new = lctx.n_outputs;
if (n_outputs_new) {
GGML_ASSERT( n_outputs_prev + n_outputs_new <= n_outputs);
GGML_ASSERT((n_outputs_prev + n_outputs_new)*n_embd <= (int64_t) lctx.embd_size);
ggml_backend_tensor_get_async(backend_embd, embd, embd_out, 0, n_outputs_new*n_embd*sizeof(float));
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
} break;
case LLAMA_POOLING_TYPE_CLS:
case LLAMA_POOLING_TYPE_MEAN:
{
GGML_ASSERT(strcmp(embd->name, "result_embd_pooled") == 0);
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// extract sequence embeddings
auto & embd_seq_out = lctx.embd_seq;
embd_seq_out.clear();
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
for (uint32_t i = 0; i < n_tokens; i++) {
const llama_seq_id seq_id = u_batch.seq_id[i][0];
if (embd_seq_out.find(seq_id) != embd_seq_out.end()) {
continue;
}
embd_seq_out[seq_id].resize(n_embd);
ggml_backend_tensor_get_async(backend_embd, embd, embd_seq_out[seq_id].data(), (n_embd*seq_id)*sizeof(float), n_embd*sizeof(float));
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
} break;
case LLAMA_POOLING_TYPE_UNSPECIFIED:
{
GGML_ASSERT(false && "unknown pooling type");
} break;
}
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
n_outputs_prev += lctx.n_outputs;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
llama : support negative ith in llama_get_ API (#6519) * llama_sampling_sample with default args is more naively usable * Batches populated by either llama_batch_get_one or llama_batch_add work with default args * Previously get_one could use the default argument * Previously add should usually have used the last index where logits[idx] == true * This hopefully encourages the use of llama_batch_add * By giving expected results when using default arguments. * Adds "negative indexing" feature to llama_get_logits_ith and llama_get_embeddings_ith * Believed to work with any currently well behaved program * Default arg now works for both cases (previously would give strange results for add case) * Any non-negative number is unaffected and behaves as previously * Negative arguments were previously invalid. * Implemented as a special case of indexing as suggested by @compilade in https://github.com/ggerganov/llama.cpp/pull/6519 * Fixed mismatch type errors * cited in macOS CI tests * Missed in original updates based on PR feedback in https://github.com/ggerganov/llama.cpp/pull/6519
2024-04-08 13:02:30 +00:00
// set to total number of outputs in the batch, for use in llama_get_logits_ith
lctx.n_outputs = n_outputs;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// wait for the computation to finish (automatically done when obtaining the model output)
//llama_synchronize(&lctx);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// decide if we need to defrag the kv cache
llama : optimize defrag moves + fix fragmentation calculation (#6037) * attempt to reduce the impact of a worst-case scenario * fragmentation calculation fix * Update llama.cpp --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-14 10:56:48 +00:00
if (cparams.causal_attn && cparams.defrag_thold >= 0.0f) {
const float fragmentation = kv_self.n >= 128 ? 1.0f - float(kv_self.used)/float(kv_self.n) : 0.0f;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// queue defragmentation for next llama_kv_cache_update
if (fragmentation > cparams.defrag_thold) {
//LLAMA_LOG_INFO("fragmentation: %.2f\n", fragmentation);
llama_kv_cache_defrag(kv_self);
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
}
Reset schedule earlier to allow overlap with ggml graph computation on device (#6933) * Reset schedule earlier to allow overlap with graph computation on device
2024-04-26 18:08:30 +00:00
// Reset state for the next token before backend sync, to allow the CPU activities in the reset to
// overlap with device computation.
ggml_backend_sched_reset(lctx.sched);
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
return 0;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
// find holes from the beginning of the KV cache and fill them by moving data from the end of the cache
static void llama_kv_cache_defrag_internal(struct llama_context & lctx) {
auto & kv_self = lctx.kv_self;
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
const auto & hparams = lctx.model.hparams;
const uint32_t n_layer = hparams.n_layer;
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
const uint32_t n_kv = llama_kv_cache_cell_max(kv_self);
const uint32_t n_used = kv_self.used;
assert(n_used <= n_kv);
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
//const int64_t t_start = ggml_time_us();
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
// number of cells moved
uint32_t n_moves = 0;
llama : optimize defrag moves + fix fragmentation calculation (#6037) * attempt to reduce the impact of a worst-case scenario * fragmentation calculation fix * Update llama.cpp --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-14 10:56:48 +00:00
// each move requires 6*n_layer tensors (see build_defrag)
// - source view, destination view, copy operation
// - x2 for keys and values
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
//const uint32_t max_moves = LLAMA_MAX_NODES/(6*n_layer);
// TODO: tmp fix https://github.com/ggerganov/llama.cpp/issues/6685#issuecomment-2057579516
const uint32_t max_moves = (LLAMA_MAX_NODES - 2*n_layer)/(6*n_layer);
llama : optimize defrag moves + fix fragmentation calculation (#6037) * attempt to reduce the impact of a worst-case scenario * fragmentation calculation fix * Update llama.cpp --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-14 10:56:48 +00:00
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
// determine which KV cells to move where
//
// cell i moves to ids[i]
//
// if ids[i] == i || ids[i] == n_kv, then cell i is not moved
//
std::vector<uint32_t> ids(n_kv, n_kv);
for (uint32_t i0 = 0; i0 < n_used; ++i0) {
const auto & cell0 = kv_self.cells[i0];
if (!cell0.is_empty()) {
ids[i0] = i0;
continue;
}
// found a hole - fill it with data from the end of the cache
uint32_t nh = 1;
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
// determine the size of the hole
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
while (i0 + nh < n_used && kv_self.cells[i0 + nh].is_empty()) {
nh++;
}
uint32_t nf = 0;
uint32_t is = n_kv - 1;
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
// starting from the end, find nh non-empty cells
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
for (; is > i0; --is) {
const auto & cell1 = kv_self.cells[is];
if (cell1.is_empty() || ids[is] != n_kv) {
continue;
}
// non-empty cell which is not yet moved
nf++;
if (nf == nh) {
break;
}
}
// this can only happen if `n_used` is not accurate, which would be a bug
GGML_ASSERT(nf == nh && "KV defrag bug: nf != nh");
nf = 0;
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
uint32_t i1 = is;
// are we moving a continuous block of memory?
bool cont = false;
llama : optimize defrag moves + fix fragmentation calculation (#6037) * attempt to reduce the impact of a worst-case scenario * fragmentation calculation fix * Update llama.cpp --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-14 10:56:48 +00:00
// should we stop searching for the next move?
bool stop = false;
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
// go back and move the nf cells to the hole
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
for (; i1 < n_kv; ++i1) {
auto & cell1 = kv_self.cells[i1];
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
if (cell1.is_empty() || ids[i1] != n_kv) {
llama : optimize defrag moves + fix fragmentation calculation (#6037) * attempt to reduce the impact of a worst-case scenario * fragmentation calculation fix * Update llama.cpp --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-14 10:56:48 +00:00
if (n_moves == max_moves) {
stop = true;
break;
}
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
cont = false;
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
continue;
}
// this cell goes to (i0 + nf)
ids[i1] = i0 + nf;
// move the cell meta data
kv_self.cells[i0 + nf] = cell1;
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
// clear the old cell and move the head there
cell1 = llama_kv_cell();
kv_self.head = n_used;
if (!cont) {
n_moves++;
cont = true;
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
nf++;
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
if (nf == nh) {
break;
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
}
llama : optimize defrag moves + fix fragmentation calculation (#6037) * attempt to reduce the impact of a worst-case scenario * fragmentation calculation fix * Update llama.cpp --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-14 10:56:48 +00:00
if (stop || n_moves == max_moves) {
break;
}
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
//LLAMA_LOG_INFO("(tmp log) KV defrag: move [%u, %u) to [%u, %u)\n", is, i1 + 1, i0, i0 + nh);
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
i0 += nh - 1;
}
if (n_moves == 0) {
return;
}
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
//LLAMA_LOG_INFO("(tmp log) KV defrag cell moves: %u\n", n_moves);
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
//LLAMA_LOG_INFO("expected gf nodes: %u\n", 6*n_moves*n_layer);
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
#if 0
// CPU defrag
//
// TODO: optimizations are possible:
// - multiple threads
// - avoid copying to the host memory when already there
//
// likely not worth the effort, as we have ggml_graph based defrag
//
const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa();
const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa();
const uint32_t kv_size = kv_self.size;
std::vector<uint8_t> buf_k;
std::vector<uint8_t> buf_v;
for (uint32_t il = 0; il < n_layer; ++il) {
const size_t k_size_row = ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa);
const size_t k_size = ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa*kv_size);
const size_t v_size_el = ggml_type_size(kv_self.v_l[il]->type);
const size_t v_size = ggml_row_size (kv_self.v_l[il]->type, n_embd_v_gqa*kv_size);
buf_k.resize(k_size);
buf_v.resize(v_size);
ggml_backend_tensor_get(kv_self.k_l[il], buf_k.data(), 0, buf_k.size());
ggml_backend_tensor_get(kv_self.v_l[il], buf_v.data(), 0, buf_v.size());
// batch move [i, i+nm) to [id, id+nm)
// note: cells can move only to a lower index
for (uint32_t i = 0; i < n_kv; ++i) {
const uint32_t id = ids[i];
if (i == id || id == n_kv) {
continue;
}
uint32_t nm = 1;
while (i + nm < n_kv && ids[i + nm] == id + nm) {
nm++;
}
// move keys
{
const int64_t os = i*k_size_row;
const int64_t od = id*k_size_row;
memcpy(buf_k.data() + od, buf_k.data() + os, nm*k_size_row);
}
// move values (note: they are transposed)
{
const int64_t os = i;
const int64_t od = id;
for (uint32_t j = 0; j < n_embd_v_gqa; ++j) {
memcpy(buf_v.data() + (od + j*kv_size)*v_size_el, buf_v.data() + (os + j*kv_size)*v_size_el, nm*v_size_el);
}
}
i += nm - 1;
}
ggml_backend_tensor_set(kv_self.k_l[il], buf_k.data(), 0, buf_k.size());
ggml_backend_tensor_set(kv_self.v_l[il], buf_v.data(), 0, buf_v.size());
}
#else
// ggml_graph defrag
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
ggml_backend_sched_reset(lctx.sched);
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ggml_cgraph * gf = llama_build_graph_defrag(lctx, ids);
llama_graph_compute(lctx, gf, lctx.cparams.n_threads);
#endif
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
//const int64_t t_end = ggml_time_us();
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
//LLAMA_LOG_INFO("(tmp log) KV defrag time: %.3f ms\n", (t_end - t_start)/1000.0);
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
}
static void llama_kv_cache_update_internal(struct llama_context & lctx) {
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
bool need_reserve = false;
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
// apply K-shift if needed
if (lctx.model.hparams.rope_type != LLAMA_ROPE_TYPE_NONE && lctx.kv_self.has_shift) {
{
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
ggml_backend_sched_reset(lctx.sched);
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
ggml_cgraph * gf = llama_build_graph_k_shift(lctx);
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
ggml_backend_sched_alloc_graph(lctx.sched, gf);
llama_set_k_shift(lctx);
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
llama_graph_compute(lctx, gf, lctx.cparams.n_threads);
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
need_reserve = true;
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
}
{
auto & kv_self = lctx.kv_self;
kv_self.has_shift = false;
for (uint32_t i = 0; i < kv_self.size; ++i) {
kv_self.cells[i].delta = 0;
}
}
}
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
if (lctx.kv_self.recurrent && lctx.kv_self.do_copy) {
{
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
ggml_backend_sched_reset(lctx.sched);
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
ggml_cgraph * gf = llama_build_graph_s_copy(lctx);
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
ggml_backend_sched_alloc_graph(lctx.sched, gf);
llama_set_s_copy(lctx);
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
llama_graph_compute(lctx, gf, lctx.cparams.n_threads);
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
need_reserve = true;
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
}
{
auto & kv_self = lctx.kv_self;
kv_self.do_copy = false;
for (uint32_t i = 0; i < kv_self.size; ++i) {
kv_self.cells[i].src = i;
}
}
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
// defragment the KV cache if needed
if (lctx.kv_self.do_defrag) {
llama_kv_cache_defrag_internal(lctx);
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
need_reserve = true;
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
lctx.kv_self.do_defrag = false;
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// reserve a worst case graph again
if (need_reserve) {
// TODO: extract to a function
// build worst-case graph
int n_tokens = (int)std::min(lctx.cparams.n_ctx, lctx.cparams.n_ubatch);
int n_past = lctx.cparams.n_ctx - n_tokens;
llama_token token = llama_token_bos(&lctx.model); // not actually used by llama_build_graph, but required to choose between token and embedding inputs graph
ggml_cgraph * gf = llama_build_graph(lctx, llama_batch_get_one(&token, n_tokens, n_past, 0), true);
// initialize scheduler with the worst-case graph
ggml_backend_sched_reset(lctx.sched);
if (!ggml_backend_sched_reserve(lctx.sched, gf)) {
LLAMA_LOG_ERROR("%s: failed to allocate compute buffers\n", __func__);
}
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
//
// tokenizer
//
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
static enum llama_vocab_type llama_vocab_get_type(const llama_vocab & vocab) {
return vocab.type;
}
static bool llama_is_normal_token(const llama_vocab & vocab, llama_token id) {
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
GGML_ASSERT(vocab.type != LLAMA_VOCAB_TYPE_NONE);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
return vocab.id_to_token[id].type == LLAMA_TOKEN_TYPE_NORMAL;
}
static bool llama_is_unknown_token(const llama_vocab & vocab, llama_token id) {
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
GGML_ASSERT(vocab.type != LLAMA_VOCAB_TYPE_NONE);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
return vocab.id_to_token[id].type == LLAMA_TOKEN_TYPE_UNKNOWN;
}
static bool llama_is_control_token(const llama_vocab & vocab, llama_token id) {
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
GGML_ASSERT(vocab.type != LLAMA_VOCAB_TYPE_NONE);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
return vocab.id_to_token[id].type == LLAMA_TOKEN_TYPE_CONTROL;
}
static bool llama_is_byte_token(const llama_vocab & vocab, llama_token id) {
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
GGML_ASSERT(vocab.type != LLAMA_VOCAB_TYPE_NONE);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
return vocab.id_to_token[id].type == LLAMA_TOKEN_TYPE_BYTE;
}
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
static bool llama_is_user_defined_token(const llama_vocab& vocab, llama_token id) {
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
GGML_ASSERT(vocab.type != LLAMA_VOCAB_TYPE_NONE);
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
return vocab.id_to_token[id].type == LLAMA_TOKEN_TYPE_USER_DEFINED;
}
static uint8_t llama_token_to_byte(const llama_vocab& vocab, llama_token id) {
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
GGML_ASSERT(llama_vocab_get_type(vocab) != LLAMA_VOCAB_TYPE_NONE);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
GGML_ASSERT(llama_is_byte_token(vocab, id));
tests : add test-tokenizer-0.sh + fix some tokenizers (#7036) * tests : add test-tokenizer-0.sh * unicode : add all unicode number ranges * starcoder : fix pre-tokenizer * tests : add test that fails with DeepSeek tokenizers * falcon : fix regex * unicode : regenerate unicode tables * refact : add tokenizer model * lint : fix * tests : disable failing tests ggml-ci * refact : add tests files ggml-ci * convert : print -> logging ggml-ci * lint : fix * unicode : digit -> number * phi-3 : update
2024-05-04 05:32:32 +00:00
const auto & token_data = vocab.id_to_token.at(id);
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
switch (llama_vocab_get_type(vocab)) {
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
case LLAMA_VOCAB_TYPE_SPM: {
auto buf = token_data.text.substr(3, 2);
return strtol(buf.c_str(), NULL, 16);
}
case LLAMA_VOCAB_TYPE_BPE: {
GGML_ASSERT(false);
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
return unicode_utf8_to_byte(token_data.text); // TODO: why is this here after GGML_ASSERT?
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
}
case LLAMA_VOCAB_TYPE_WPM: {
GGML_ASSERT(false);
}
default:
GGML_ASSERT(false);
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
static llama_token llama_byte_to_token(const llama_vocab & vocab, uint8_t ch) {
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
GGML_ASSERT(llama_vocab_get_type(vocab) != LLAMA_VOCAB_TYPE_NONE);
llama : validate special token ids are in range when loading GGUF model (#3635) * Add validation for special token ids to llama.cpp Small optimization for llama_byte_to_token SPM mode * Fix BPE newline check, only I could break something so simple * Killll meeeeee * Account for GGUF_KEY_KEY only setting when the key exists * Minor code cleanups. * Fix convert.py error msg when added tokens are out of range * Make gguf SpecialVocab vocab size-aware Update conversion scripts accordingly * Avoid a string copy Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-22 18:14:56 +00:00
static const char * hex = "0123456789ABCDEF";
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
switch (llama_vocab_get_type(vocab)) {
llama : check LLAMA_TRACE env for extra logging (#4929) * llama : minor fix indent * llama : check LLAMA_TRACE env for extra logging ggml-ci
2024-01-14 11:26:53 +00:00
case LLAMA_VOCAB_TYPE_SPM: {
const char buf[7] = { '<', '0', 'x', hex[ch >> 4], hex[ch & 15], '>', 0 };
llama : allow raw byte in SPM vocabs; don't crash on nl 404 (#5478) * common : don't crash if newline token is not found * common : llama_byte_to_token: allow falling back to finding just the token byte in SPM vocabs
2024-02-13 16:18:16 +00:00
auto token = vocab.token_to_id.find(buf);
if (token != vocab.token_to_id.end()) {
return (*token).second;
}
// Try to fall back to just the byte as a string
const char buf2[2] = { (char)ch, 0 };
return vocab.token_to_id.at(buf2);
llama : check LLAMA_TRACE env for extra logging (#4929) * llama : minor fix indent * llama : check LLAMA_TRACE env for extra logging ggml-ci
2024-01-14 11:26:53 +00:00
}
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
case LLAMA_VOCAB_TYPE_WPM:
llama : check LLAMA_TRACE env for extra logging (#4929) * llama : minor fix indent * llama : check LLAMA_TRACE env for extra logging ggml-ci
2024-01-14 11:26:53 +00:00
case LLAMA_VOCAB_TYPE_BPE: {
llama : refactor unicode stuff (#5992) * llama : refactor unicode stuff ggml-ci * unicode : names * make : fix c++ compiler * unicode : names * unicode : straighten tables * zig : fix build * unicode : put nfd normalization behind API ggml-ci * swift : fix build * unicode : add BOM * unicode : add <cstdint> ggml-ci * unicode : pass as cpts as const ref
2024-03-11 15:47:47 +00:00
return vocab.token_to_id.at(unicode_byte_to_utf8(ch));
llama : check LLAMA_TRACE env for extra logging (#4929) * llama : minor fix indent * llama : check LLAMA_TRACE env for extra logging ggml-ci
2024-01-14 11:26:53 +00:00
}
default:
GGML_ASSERT(false);
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llama : more tokenizer fixes (#2810) * tests : write a Python tokenizer test (wip) * llama : prefix input text for tokenization with whitespace * llama : distinguish pieces from decoded text + fix detokenization * common : add comments * examples : no longer manually add leading space when tokenizing * tests : use Python to generate tokenizer tests for C++ * tests : add option to tokenize text files ggml-ci * tests : add test-tokenizer-1.py * llama.cpp : fix LF token * hellaswag : move the concat space for clarity * tests : add falcon tests (py + cpp, currently do not pass Unicode) ggml-ci * common : temporary separate llama_detokenize calls for SPM and BPE --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com>
2023-08-27 11:19:19 +00:00
static void llama_escape_whitespace(std::string & text) {
replace_all(text, " ", "\xe2\x96\x81");
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llama : escape all U+2581 in a string (#2750)
2023-08-24 09:26:01 +00:00
static void llama_unescape_whitespace(std::string & word) {
replace_all(word, "\xe2\x96\x81", " ");
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
struct llm_symbol {
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
using index = int;
index prev;
index next;
const char * text;
size_t n;
};
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
static_assert(std::is_trivially_copyable<llm_symbol>::value, "llm_symbol is not trivially copyable");
// SPM tokenizer
// original implementation:
// https://github.com/ggerganov/llama.cpp/commit/074bea2eb1f1349a0118239c4152914aecaa1be4
llama : fix various warnings
2023-05-13 08:23:15 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
struct llm_bigram_spm {
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
struct comparator {
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
bool operator()(llm_bigram_spm & l, llm_bigram_spm & r) {
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
return (l.score < r.score) || (l.score == r.score && l.left > r.left);
}
};
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
using queue_storage = std::vector<llm_bigram_spm>;
using queue = std::priority_queue<llm_bigram_spm, queue_storage, comparator>;
llm_symbol::index left;
llm_symbol::index right;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
float score;
size_t size;
};
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
struct llm_tokenizer_spm {
tests : multi-thread the tokenizer tests (#5474) * tests : multi-thread the tokenizer tests ggml-ci * unicode : fix data race for unidentified codepoints ggml-ci * unicode : minor style fixes ggml-ci
2024-02-13 13:14:22 +00:00
llm_tokenizer_spm(const llama_vocab & vocab) : vocab(vocab) {}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
void tokenize(const std::string & text, std::vector<llama_vocab::id> & output) {
// split string into utf8 chars
int index = 0;
size_t offs = 0;
while (offs < text.size()) {
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llm_symbol sym;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
size_t len = utf8_len(text[offs]);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
sym.text = text.c_str() + offs;
whisper : tokenizer fix + re-enable tokenizer test for LLaMa (#3096) * Fix für #2721 * Reenable tokenizer test for LLaMa * Add `console.cpp` dependency * Fix dependency to `common` * Fixing wrong fix. * Make console usage platform specific Work on compiler warnings. * Adapting makefile * Remove trailing whitespace * Adapting the other parts of the makefile * Fix typo.
2023-09-13 13:19:44 +00:00
sym.n = std::min(len, text.size() - offs);
offs += sym.n;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
sym.prev = index - 1;
sym.next = offs == text.size() ? -1 : index + 1;
index++;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
symbols.emplace_back(sym);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
// seed the work queue with all possible 2-character tokens.
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
for (size_t i = 1; i < symbols.size(); ++i) {
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
try_add_bigram(i - 1, i);
}
// keep substituting the highest frequency pairs for as long as we can.
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
while (!work_queue.empty()) {
auto bigram = work_queue.top();
work_queue.pop();
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
auto & left_sym = symbols[bigram.left];
auto & right_sym = symbols[bigram.right];
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
// if one of the symbols already got merged, skip it.
if (left_sym.n == 0 || right_sym.n == 0 ||
left_sym.n + right_sym.n != bigram.size) {
continue;
}
// merge the right sym into the left one
left_sym.n += right_sym.n;
right_sym.n = 0;
add log_callback to llama_context_params for custom logging. (#2234) * add log_callback to llama_context_params for custom logging. * Fix macro expansion on gcc * Add struct llama_state for global variables and move log_callback there * Turn log level into enum and some minor changes. * Remove model_for_logging parameter (not needed anymore) * Convert remaining fprintf(stderr, ...) calls to use new macros. * Fix enum and initialize g_state * Fix log calls after merge * Fix missing static * Add back all the new lines in the logging strings * Add comment for llama_log_callback and replace remaining printf calls --------- Co-authored-by: grahameth <-> Co-authored-by: Helmut <helmut.buhler@inf.h-brs.de>
2023-08-09 20:46:40 +00:00
//LLAMA_LOG_INFO("left = '%*s' size = %zu\n", (int) left_sym.n, left_sym.text, bigram.size);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
// remove the right sym from the chain
left_sym.next = right_sym.next;
if (right_sym.next >= 0) {
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
symbols[right_sym.next].prev = bigram.left;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
// find more substitutions
try_add_bigram(left_sym.prev, bigram.left);
try_add_bigram(bigram.left, left_sym.next);
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
for (int i = 0; i != -1; i = symbols[i].next) {
auto & symbol = symbols[i];
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
resegment(symbol, output);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
}
private:
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
void resegment(llm_symbol & symbol, std::vector<llama_vocab::id> & output) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
auto text = std::string(symbol.text, symbol.n);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
auto token = vocab.token_to_id.find(text);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
// Do we need to support is_unused?
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
if (token != vocab.token_to_id.end()) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
output.push_back((*token).second);
return;
}
const auto p = rev_merge.find(text);
if (p == rev_merge.end()) {
// output any symbols that did not form tokens as bytes.
tests : multi-thread the tokenizer tests (#5474) * tests : multi-thread the tokenizer tests ggml-ci * unicode : fix data race for unidentified codepoints ggml-ci * unicode : minor style fixes ggml-ci
2024-02-13 13:14:22 +00:00
output.reserve(output.size() + symbol.n);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
for (int j = 0; j < (int)symbol.n; ++j) {
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama_vocab::id token_id = llama_byte_to_token(vocab, symbol.text[j]);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
output.push_back(token_id);
}
return;
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
resegment(symbols[p->second.first], output);
resegment(symbols[p->second.second], output);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
void try_add_bigram(int left, int right) {
if (left == -1 || right == -1) {
return;
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
const std::string text = std::string(symbols[left].text, symbols[left].n + symbols[right].n);
auto token = vocab.token_to_id.find(text);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
if (token == vocab.token_to_id.end()) {
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
return;
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
if (static_cast<size_t>((*token).second) >= vocab.id_to_token.size()) {
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
return;
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
const auto & tok_data = vocab.id_to_token[(*token).second];
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llm_bigram_spm bigram;
bigram.left = left;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
bigram.right = right;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
bigram.score = tok_data.score;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
bigram.size = text.size();
work_queue.push(bigram);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
// Do we need to support is_unused?
rev_merge[text] = std::make_pair(left, right);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
const llama_vocab & vocab;
std::vector<llm_symbol> symbols;
llm_bigram_spm::queue work_queue;
std::map<std::string, std::pair<int, int>> rev_merge;
};
// BPE tokenizer
// adapted from https://github.com/cmp-nct/ggllm.cpp [MIT License]
// tried to simplify unicode stuff, so most likely does not work 100% correctly!
// TODO: there are a lot of common parts between spm and bpe tokenizers, should be refactored and reused
struct llm_bigram_bpe {
struct comparator {
10X faster BPE tokenizer (#2876) * 10X faster BPE tokenizer * Remove comment that no longer applies --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2023-08-29 20:55:03 +00:00
bool operator()(const llm_bigram_bpe & l, const llm_bigram_bpe & r) const {
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
return l.rank > r.rank || (l.rank == r.rank && l.left > r.left);
}
};
using queue_storage = std::vector<llm_bigram_bpe>;
using queue = std::priority_queue<llm_bigram_bpe, queue_storage, comparator>;
llm_symbol::index left;
llm_symbol::index right;
std::string text;
int rank;
size_t size;
};
struct llm_tokenizer_bpe {
Fix spm whitespaces (#2806) * llama.cpp : fix spm whitespace escaping + clean up * main.cpp : spm - add whitespace in front of prompt * test-tokenizer-0.cpp : spm - add whitespace in front of prompt
2023-08-26 11:45:53 +00:00
llm_tokenizer_bpe(const llama_vocab & vocab): vocab(vocab) {}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
void tokenize(const std::string & text, std::vector<llama_vocab::id> & output) {
int final_prev_index = -1;
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
std::vector<std::string> word_collection;
switch (vocab.type) {
case LLAMA_VOCAB_TYPE_BPE:
switch (vocab.type_pre) {
case LLAMA_VOCAB_PRE_TYPE_LLAMA3:
word_collection = unicode_regex_split(text, {
// original regex from tokenizer.json
//"(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?\\p{L}+|\\p{N}{1,3}| ?[^\\s\\p{L}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
// adapted: https://github.com/ggerganov/llama.cpp/pull/6920#issuecomment-2080233989
"(?:'[sS]|'[tT]|'[rR][eE]|'[vV][eE]|'[mM]|'[lL][lL]|'[dD])|[^\\r\\n\\p{L}\\p{N}]?\\p{L}+|\\p{N}{1,3}| ?[^\\s\\p{L}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
});
break;
case LLAMA_VOCAB_PRE_TYPE_DEEPSEEK_LLM:
word_collection = unicode_regex_split(text, {
"[\r\n]",
"\\s?[A-Za-zµÀ-ÖØ-öø-ƺƼ-ƿDŽ-ʓʕ-ʯͰ-ͳͶͷͻ-ͽͿΆΈ-ΊΌΎ-ΡΣ-ϵϷ-ҁҊ-ԯԱ-ՖႠ-ჅᎠ-Ᏽᏸ-ᏽᲐ-ᲺᲽ-Ჿᴀ-ᴫᵫ-ᵷᵹ-ᶚḀ-ἕἘ-Ἕἠ-ὅὈ-Ὅὐ-ὗὙὛὝὟ-ώᾀ-ᾴᾶ-ᾼιῂ-ῄῆ-ῌῐ-ΐῖ-Ίῠ-Ῥῲ-ῴῶ-ῼℂℇℊ-ℓℕℙ-ℝℤΩℨK-ℭℯ-ℴℹℼ-ℿⅅ-ⅉⅎↃↄⰀ-ⱻⱾ-ⳤⳫ-ⳮⳲⳳꙀ-ꙭꚀ-ꚛꜢ-ꝯꝱ-ꞇꞋ-ꞎꭰ-ꮿff-stﬓ-ﬗＡ-Ｚａ-ｚ𐐀-𐑏𐒰-𐓓𐓘-𐓻𐲀-𐲲𐳀-𐳲𑢠-𑣟𞤀-𞥃]+",
"\\s?[!-/:-~！-／：-～‘-‟　-。]+",
"\\s+$",
"[一-龥ࠀ-一가-퟿]+",
"\\p{N}+",
});
break;
case LLAMA_VOCAB_PRE_TYPE_DEEPSEEK_CODER:
word_collection = unicode_regex_split(text, {
"[\r\n]",
"\\s?\\p{L}+",
"\\s?\\p{P}+",
"[一-龥ࠀ-一가-퟿]+",
tests : add test-tokenizer-0.sh + fix some tokenizers (#7036) * tests : add test-tokenizer-0.sh * unicode : add all unicode number ranges * starcoder : fix pre-tokenizer * tests : add test that fails with DeepSeek tokenizers * falcon : fix regex * unicode : regenerate unicode tables * refact : add tokenizer model * lint : fix * tests : disable failing tests ggml-ci * refact : add tests files ggml-ci * convert : print -> logging ggml-ci * lint : fix * unicode : digit -> number * phi-3 : update
2024-05-04 05:32:32 +00:00
"\\p{N}",
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
});
break;
case LLAMA_VOCAB_PRE_TYPE_FALCON:
word_collection = unicode_regex_split(text, {
"[\\p{P}\\$\\+<=>\\^~\\|]+",
"'s|'t|'re|'ve|'m|'ll|'d| ?\\p{L}+| ?\\p{N}+| ?[^\\s\\p{L}\\p{N}]+|\\s+(?!\\S)",
"[0-9][0-9][0-9]",
});
break;
case LLAMA_VOCAB_PRE_TYPE_MPT:
// TODO: MPT pre-tokenization regexes are unknown
// the following are close, but not exact. run the following:
// ./bin/test-tokenizer-0 ../models/ggml-vocab-mpt.gguf
GGML_ASSERT("MPT pre-tokenization regexes are unknown - fixes needed");
word_collection = unicode_regex_split(text, {
"\\s?\\p{L}+",
"\\s?\\p{P}+",
"'s|'t|'re|'ve|'m|'ll|'d| ?\\p{L}+| ?\\p{N}+| ?[^\\s\\p{L}\\p{N}]+|\\s+(?!\\S)",
});
break;
case LLAMA_VOCAB_PRE_TYPE_STARCODER:
tests : add test-tokenizer-0.sh + fix some tokenizers (#7036) * tests : add test-tokenizer-0.sh * unicode : add all unicode number ranges * starcoder : fix pre-tokenizer * tests : add test that fails with DeepSeek tokenizers * falcon : fix regex * unicode : regenerate unicode tables * refact : add tokenizer model * lint : fix * tests : disable failing tests ggml-ci * refact : add tests files ggml-ci * convert : print -> logging ggml-ci * lint : fix * unicode : digit -> number * phi-3 : update
2024-05-04 05:32:32 +00:00
case LLAMA_VOCAB_PRE_TYPE_REFACT:
command-r : add BPE pre-tokenization (#7063) * Add BPE pre-tokenization for Command-R/R+. * Bump transformers convert requirement. * command-r : add individual digits regex --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-05-05 05:19:30 +00:00
case LLAMA_VOCAB_PRE_TYPE_COMMAND_R:
tests : add test-tokenizer-0.sh + fix some tokenizers (#7036) * tests : add test-tokenizer-0.sh * unicode : add all unicode number ranges * starcoder : fix pre-tokenizer * tests : add test that fails with DeepSeek tokenizers * falcon : fix regex * unicode : regenerate unicode tables * refact : add tokenizer model * lint : fix * tests : disable failing tests ggml-ci * refact : add tests files ggml-ci * convert : print -> logging ggml-ci * lint : fix * unicode : digit -> number * phi-3 : update
2024-05-04 05:32:32 +00:00
word_collection = unicode_regex_split(text, {
"\\p{N}",
"'s|'t|'re|'ve|'m|'ll|'d| ?\\p{L}+| ?\\p{N}+| ?[^\\s\\p{L}\\p{N}]+|\\s+(?!\\S)",
});
break;
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
case LLAMA_VOCAB_PRE_TYPE_GPT2:
Fix OLMo HF to GGUF conversion (#6910)
2024-05-07 19:39:43 +00:00
case LLAMA_VOCAB_PRE_TYPE_OLMO:
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
word_collection = unicode_regex_split(text, {
"'s|'t|'re|'ve|'m|'ll|'d| ?\\p{L}+| ?\\p{N}+| ?[^\\s\\p{L}\\p{N}]+|\\s+(?!\\S)",
});
break;
default:
// default regex for BPE tokenization pre-processing
word_collection = unicode_regex_split(text, {
"[\\p{P}\\$\\+<=>\\^~\\|]+",
"'s|'t|'re|'ve|'m|'ll|'d| ?\\p{L}+| ?\\p{N}+| ?[^\\s\\p{L}\\p{N}]+|\\s+(?!\\S)",
"\\p{N}+",
"[0-9][0-9][0-9]",
});
break;
}
break;
default:
GGML_ASSERT(false);
break;
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
symbols_final.clear();
for (auto & word : word_collection) {
work_queue = llm_bigram_bpe::queue();
symbols.clear();
int index = 0;
size_t offset = 0;
while (offset < word.size()) {
llm_symbol sym;
size_t char_len = std::min(word.size() - offset, (size_t) ::utf8_len(word[offset]));
sym.text = word.c_str() + offset;
sym.n = char_len;
offset += sym.n;
sym.prev = index - 1;
sym.next = offset == word.size() ? -1 : index + 1;
index++;
symbols.emplace_back(sym);
}
for (size_t i = 1; i < symbols.size(); ++i) {
add_new_bigram(i - 1, i);
}
// build token(s)
while (!work_queue.empty()) {
auto bigram = work_queue.top();
work_queue.pop();
auto & left_symbol = symbols[bigram.left];
auto & right_symbol = symbols[bigram.right];
if (left_symbol.n == 0 || right_symbol.n == 0) {
continue;
}
std::string left_token = std::string(left_symbol.text, left_symbol.n);
std::string right_token = std::string(right_symbol.text, right_symbol.n);
if (left_token + right_token != bigram.text) {
continue; // Skip this bigram if it's outdated
}
// merge the right sym into the left one
left_symbol.n += right_symbol.n;
right_symbol.n = 0;
// remove the right sym from the chain
left_symbol.next = right_symbol.next;
if (right_symbol.next >= 0) {
symbols[right_symbol.next].prev = bigram.left;
}
add_new_bigram(left_symbol.prev, bigram.left); // left side of current symbol
add_new_bigram(bigram.left, left_symbol.next); // right side of current symbol
}
eval-callback: Example how to use eval callback for debugging (#6576) * gguf-debug: Example how to use ggml callback for debugging * gguf-debug: no mutex, verify type, fix stride. * llama: cv eval: move cb eval field in common gpt_params * ggml_debug: use common gpt_params to pass cb eval. Fix get tensor SIGV random. * ggml_debug: ci: add tests * ggml_debug: EOL in CMakeLists.txt * ggml_debug: Remove unused param n_batch, no batching here * ggml_debug: fix trailing spaces * ggml_debug: fix trailing spaces * common: fix cb_eval and user data not initialized * ci: build revert label * ggml_debug: add main test label * doc: add a model: add a link to ggml-debug * ggml-debug: add to make toolchain * ggml-debug: tests add the main label * ggml-debug: ci add test curl label * common: allow the warmup to be disabled in llama_init_from_gpt_params * ci: add curl test * ggml-debug: better tensor type support * gitignore : ggml-debug * ggml-debug: printing also the sum of each tensor * ggml-debug: remove block size * eval-callback: renamed from ggml-debug * eval-callback: fix make toolchain --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-11 12:51:07 +00:00
// add the finished tokens to the final list keeping correct order for next and prev
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
for (auto & sym : symbols) {
if (sym.n > 0) {
sym.prev = final_prev_index;
sym.next = -1;
if (final_prev_index != -1) {
symbols_final[final_prev_index].next = symbols_final.size();
}
symbols_final.emplace_back(sym);
final_prev_index = symbols_final.size() - 1;
}
}
}
symbols = symbols_final;
if (!symbols.empty()) {
for (int i = 0; i != -1; i = symbols[i].next) {
auto & symbol = symbols[i];
if (symbol.n == 0) {
continue;
}
const std::string str = std::string(symbol.text, symbol.n);
const auto token = vocab.token_to_id.find(str);
if (token == vocab.token_to_id.end()) {
for (auto j = str.begin(); j != str.end(); ++j) {
std::string byte_str(1, *j);
auto token_multibyte = vocab.token_to_id.find(byte_str);
if (token_multibyte == vocab.token_to_id.end()) {
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
throw std::runtime_error("ERROR: byte not found in vocab");
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
output.push_back((*token_multibyte).second);
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
}
} else {
output.push_back((*token).second);
}
}
}
}
private:
void add_new_bigram(int left, int right) {
if (left == -1 || right == -1) {
return;
}
std::string left_token = std::string(symbols[left].text, symbols[left].n);
std::string right_token = std::string(symbols[right].text, symbols[right].n);
int rank_found = -1;
rank_found = vocab.find_bpe_rank(left_token, right_token);
if (rank_found < 0) {
return;
}
llm_bigram_bpe bigram;
bigram.left = left;
bigram.right = right;
bigram.text = left_token + right_token;
bigram.size = left_token.size() + right_token.size();
bigram.rank = rank_found;
work_queue.push(bigram);
}
const llama_vocab & vocab;
std::vector<llm_symbol> symbols;
std::vector<llm_symbol> symbols_final;
llm_bigram_bpe::queue work_queue;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
};
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
struct llm_tokenizer_wpm {
llm_tokenizer_wpm(const llama_vocab & vocab): vocab(vocab) {}
void tokenize(const std::string & text, std::vector<llama_vocab::id> & output) {
auto * token_map = &vocab.token_to_id;
// normalize and split by whitespace
std::vector<std::string> words = preprocess(text);
// bos token prepended already
// find the longest tokens that form the words
for (const std::string &word : words) {
// skip empty words
if (word.size() == 0) {
continue;
}
// prepend phantom space
std::string word1 = "\xe2\x96\x81" + word;
int n = word1.size();
// we're at the start of a new word
int i = 0;
bool match_any = false;
// move through character position in word
while (i < n) {
// loop through possible match length
bool match = false;
for (int j = n; j > i; j--) {
auto it = token_map->find(word1.substr(i, j - i));
if (it != token_map->end()) {
output.push_back(it->second);
match = true;
match_any = true;
i = j;
break;
}
}
// must be an unknown character
if (!match) {
i++;
}
}
// we didn't find any matches for this word
if (!match_any) {
output.push_back(vocab.special_unk_id);
}
}
}
std::vector<std::string> preprocess(const std::string & text) {
llama : refactor unicode stuff (#5992) * llama : refactor unicode stuff ggml-ci * unicode : names * make : fix c++ compiler * unicode : names * unicode : straighten tables * zig : fix build * unicode : put nfd normalization behind API ggml-ci * swift : fix build * unicode : add BOM * unicode : add <cstdint> ggml-ci * unicode : pass as cpts as const ref
2024-03-11 15:47:47 +00:00
std::vector<uint32_t> cpts_nfd = unicode_cpts_normalize_nfd(unicode_cpts_from_utf8(text));
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
llama : improve BERT tokenization (#5740) * implement nfd for stripping accents in wpm tokenizer * sort nfd map; reuse iterator * use builtin tolower * add locale include * Simplify to_lower cases Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-02-28 08:51:11 +00:00
// strip accents, strip control, uniformize whitespace,
// to lowercase, pad chinese characters, pad punctuation
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
std::string new_str = "";
llama : refactor unicode stuff (#5992) * llama : refactor unicode stuff ggml-ci * unicode : names * make : fix c++ compiler * unicode : names * unicode : straighten tables * zig : fix build * unicode : put nfd normalization behind API ggml-ci * swift : fix build * unicode : add BOM * unicode : add <cstdint> ggml-ci * unicode : pass as cpts as const ref
2024-03-11 15:47:47 +00:00
for (uint32_t code : cpts_nfd) {
int type = unicode_cpt_type(code);
llama : improve BERT tokenization (#5740) * implement nfd for stripping accents in wpm tokenizer * sort nfd map; reuse iterator * use builtin tolower * add locale include * Simplify to_lower cases Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-02-28 08:51:11 +00:00
if (type == CODEPOINT_TYPE_ACCENT_MARK || type == CODEPOINT_TYPE_CONTROL) {
continue;
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
}
wpm : portable unicode tolower (#6305) Also use C locale for ispunct/isspace, and split unicode-data.cpp from unicode.cpp.
2024-03-26 21:46:21 +00:00
code = unicode_tolower(code);
llama : improve BERT tokenization (#5740) * implement nfd for stripping accents in wpm tokenizer * sort nfd map; reuse iterator * use builtin tolower * add locale include * Simplify to_lower cases Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-02-28 08:51:11 +00:00
if (type == CODEPOINT_TYPE_WHITESPACE) {
code = ' ';
}
llama : refactor unicode stuff (#5992) * llama : refactor unicode stuff ggml-ci * unicode : names * make : fix c++ compiler * unicode : names * unicode : straighten tables * zig : fix build * unicode : put nfd normalization behind API ggml-ci * swift : fix build * unicode : add BOM * unicode : add <cstdint> ggml-ci * unicode : pass as cpts as const ref
2024-03-11 15:47:47 +00:00
std::string s = unicode_cpt_to_utf8(code);
llama : improve BERT tokenization (#5740) * implement nfd for stripping accents in wpm tokenizer * sort nfd map; reuse iterator * use builtin tolower * add locale include * Simplify to_lower cases Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-02-28 08:51:11 +00:00
if (type == CODEPOINT_TYPE_PUNCTUATION || is_ascii_punct(code) || is_chinese_char(code)) {
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
new_str += " ";
llama : improve BERT tokenization (#5740) * implement nfd for stripping accents in wpm tokenizer * sort nfd map; reuse iterator * use builtin tolower * add locale include * Simplify to_lower cases Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-02-28 08:51:11 +00:00
new_str += s;
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
new_str += " ";
llama : improve BERT tokenization (#5740) * implement nfd for stripping accents in wpm tokenizer * sort nfd map; reuse iterator * use builtin tolower * add locale include * Simplify to_lower cases Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-02-28 08:51:11 +00:00
} else {
new_str += s;
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
}
}
// split by whitespace
uint64_t l = 0;
uint64_t r = 0;
llama : improve BERT tokenization (#5740) * implement nfd for stripping accents in wpm tokenizer * sort nfd map; reuse iterator * use builtin tolower * add locale include * Simplify to_lower cases Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-02-28 08:51:11 +00:00
std::vector<std::string> words;
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
while (r < new_str.size()) {
// if is whitespace
wpm : portable unicode tolower (#6305) Also use C locale for ispunct/isspace, and split unicode-data.cpp from unicode.cpp.
2024-03-26 21:46:21 +00:00
if (isspace(new_str[r], std::locale::classic())) {
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
if (r > l) words.push_back(new_str.substr(l, (r - l)));
l = r + 1;
r = l;
llama : refactor unicode stuff (#5992) * llama : refactor unicode stuff ggml-ci * unicode : names * make : fix c++ compiler * unicode : names * unicode : straighten tables * zig : fix build * unicode : put nfd normalization behind API ggml-ci * swift : fix build * unicode : add BOM * unicode : add <cstdint> ggml-ci * unicode : pass as cpts as const ref
2024-03-11 15:47:47 +00:00
} else {
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
r += 1;
}
}
if (r > l) {
words.push_back(new_str.substr(l, (r - l)));
}
return words;
}
llama : improve BERT tokenization (#5740) * implement nfd for stripping accents in wpm tokenizer * sort nfd map; reuse iterator * use builtin tolower * add locale include * Simplify to_lower cases Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-02-28 08:51:11 +00:00
bool is_ascii_punct(uint32_t code) {
wpm : portable unicode tolower (#6305) Also use C locale for ispunct/isspace, and split unicode-data.cpp from unicode.cpp.
2024-03-26 21:46:21 +00:00
if (code > 0xFF) {
return false;
}
auto c = char(static_cast<unsigned char>(code));
return ispunct(c, std::locale::classic());
llama : improve BERT tokenization (#5740) * implement nfd for stripping accents in wpm tokenizer * sort nfd map; reuse iterator * use builtin tolower * add locale include * Simplify to_lower cases Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> --------- Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-02-28 08:51:11 +00:00
}
llama : refactor unicode stuff (#5992) * llama : refactor unicode stuff ggml-ci * unicode : names * make : fix c++ compiler * unicode : names * unicode : straighten tables * zig : fix build * unicode : put nfd normalization behind API ggml-ci * swift : fix build * unicode : add BOM * unicode : add <cstdint> ggml-ci * unicode : pass as cpts as const ref
2024-03-11 15:47:47 +00:00
bool is_chinese_char(uint32_t cpt) {
if ((cpt >= 0x4E00 && cpt <= 0x9FFF) ||
(cpt >= 0x3400 && cpt <= 0x4DBF) ||
(cpt >= 0x20000 && cpt <= 0x2A6DF) ||
(cpt >= 0x2A700 && cpt <= 0x2B73F) ||
(cpt >= 0x2B740 && cpt <= 0x2B81F) ||
(cpt >= 0x2B920 && cpt <= 0x2CEAF) || // this should be 0x2B820 but in hf rust code it is 0x2B920
(cpt >= 0xF900 && cpt <= 0xFAFF) ||
(cpt >= 0x2F800 && cpt <= 0x2FA1F) ||
(cpt >= 0x3000 && cpt <= 0x303F) ||
(cpt >= 0xFF00 && cpt <= 0xFFEF)) {
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
return true; // NOLINT
}
return false;
}
const llama_vocab & vocab;
};
typedef enum FRAGMENT_BUFFER_VARIANT_TYPE {
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
FRAGMENT_BUFFER_VARIANT_TYPE_TOKEN,
FRAGMENT_BUFFER_VARIANT_TYPE_RAW_TEXT
} FRAGMENT_BUFFER_VARIANT_TYPE;
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
struct fragment_buffer_variant {
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
fragment_buffer_variant(llama_vocab::id _token)
:
type(FRAGMENT_BUFFER_VARIANT_TYPE_TOKEN),
token(_token),
raw_text(_dummy),
offset(0),
tests : multi-thread the tokenizer tests (#5474) * tests : multi-thread the tokenizer tests ggml-ci * unicode : fix data race for unidentified codepoints ggml-ci * unicode : minor style fixes ggml-ci
2024-02-13 13:14:22 +00:00
length(0) {}
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
fragment_buffer_variant(const std::string & _raw_text, int64_t _offset, int64_t _length)
:
type(FRAGMENT_BUFFER_VARIANT_TYPE_RAW_TEXT),
tests : multi-thread the tokenizer tests (#5474) * tests : multi-thread the tokenizer tests ggml-ci * unicode : fix data race for unidentified codepoints ggml-ci * unicode : minor style fixes ggml-ci
2024-02-13 13:14:22 +00:00
token((llama_vocab::id) - 1),
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
raw_text(_raw_text),
offset(_offset),
length(_length){
tests : multi-thread the tokenizer tests (#5474) * tests : multi-thread the tokenizer tests ggml-ci * unicode : fix data race for unidentified codepoints ggml-ci * unicode : minor style fixes ggml-ci
2024-02-13 13:14:22 +00:00
GGML_ASSERT(_offset >= 0);
GGML_ASSERT(_length >= 1);
GGML_ASSERT(offset + length <= raw_text.length());
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
}
const FRAGMENT_BUFFER_VARIANT_TYPE type;
const llama_vocab::id token;
const std::string _dummy;
const std::string & raw_text;
const uint64_t offset;
const uint64_t length;
};
// #define PRETOKENIZERDEBUG
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
static void tokenizer_st_partition(const llama_vocab & vocab, std::forward_list<fragment_buffer_variant> & buffer) {
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
// for each special token
for (const auto & st: vocab.special_tokens_cache) {
const auto & special_token = st.first;
const auto & special_id = st.second;
// for each text fragment
std::forward_list<fragment_buffer_variant>::iterator it = buffer.begin();
while (it != buffer.end()) {
auto & fragment = (*it);
// if a fragment is text ( not yet processed )
if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_RAW_TEXT) {
auto * raw_text = &(fragment.raw_text);
auto raw_text_base_offset = fragment.offset;
auto raw_text_base_length = fragment.length;
// loop over the text
while (true) {
english : use `typos` to fix comments and logs (#4354)
2023-12-12 09:53:36 +00:00
// find the first occurrence of a given special token in this fragment
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
// passing offset argument only limit the "search area" but match coordinates
// are still relative to the source full raw_text
auto match = raw_text->find(special_token, raw_text_base_offset);
english : use `typos` to fix comments and logs (#4354)
2023-12-12 09:53:36 +00:00
// no occurrences found, stop processing this fragment for a given special token
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
if (match == std::string::npos) break;
// check if match is within bounds of offset <-> length
if (match + special_token.length() > raw_text_base_offset + raw_text_base_length) break;
#ifdef PRETOKENIZERDEBUG
llama : use LLAMA_LOG_ macros for logging
2024-01-14 09:03:19 +00:00
LLAMA_LOG_WARN("FF: (%ld %ld %ld) '%s'\n", raw_text->length(), raw_text_base_offset, raw_text_base_length, raw_text->substr(raw_text_base_offset, raw_text_base_length).c_str());
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
#endif
auto source = std::distance(buffer.begin(), it);
// if match is further than base offset
// then we have some text to the left of it
if (match > raw_text_base_offset) {
// left
const int64_t left_reminder_offset = raw_text_base_offset + 0;
const int64_t left_reminder_length = match - raw_text_base_offset;
buffer.emplace_after(it, (*raw_text), left_reminder_offset, left_reminder_length);
#ifdef PRETOKENIZERDEBUG
llama : use LLAMA_LOG_ macros for logging
2024-01-14 09:03:19 +00:00
LLAMA_LOG_WARN("FL: (%ld %ld) '%s'\n", left_reminder_offset, left_reminder_length, raw_text->substr(left_reminder_offset, left_reminder_length).c_str());
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
#endif
it++;
}
// special token
buffer.emplace_after(it, special_id);
it++;
// right
if (match + special_token.length() < raw_text_base_offset + raw_text_base_length) {
const int64_t right_reminder_offset = match + special_token.length();
const int64_t right_reminder_length = raw_text_base_length - ((match - raw_text_base_offset) + special_token.length());
buffer.emplace_after(it, (*raw_text), right_reminder_offset, right_reminder_length);
#ifdef PRETOKENIZERDEBUG
llama : use LLAMA_LOG_ macros for logging
2024-01-14 09:03:19 +00:00
LLAMA_LOG_WARN("FR: (%ld %ld) '%s'\n", right_reminder_offset, right_reminder_length, raw_text->substr(right_reminder_offset, right_reminder_length).c_str());
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
#endif
it++;
if (source == 0) {
buffer.erase_after(buffer.before_begin());
} else {
buffer.erase_after(std::next(buffer.begin(), (source-1)));
}
// repeat for the right side
raw_text_base_offset = right_reminder_offset;
raw_text_base_length = right_reminder_length;
#ifdef PRETOKENIZERDEBUG
llama : use LLAMA_LOG_ macros for logging
2024-01-14 09:03:19 +00:00
LLAMA_LOG_WARN("RR: (%ld %ld) '%s'\n", raw_text_base_offset, raw_text_base_length, raw_text->substr(raw_text_base_offset, raw_text_base_length).c_str());
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
#endif
} else {
if (source == 0) {
buffer.erase_after(buffer.before_begin());
} else {
buffer.erase_after(std::next(buffer.begin(), (source-1)));
}
break;
}
}
}
it++;
}
}
}
BERT tokenizer fixes (#6498) Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
2024-04-09 17:44:08 +00:00
static std::vector<llama_vocab::id> llama_tokenize_internal(const llama_vocab & vocab, std::string raw_text, bool add_special, bool parse_special) {
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
std::vector<llama_vocab::id> output;
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
std::forward_list<fragment_buffer_variant> fragment_buffer;
BERT tokenizer fixes (#6498) Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
2024-04-09 17:44:08 +00:00
if (!raw_text.empty()) {
fragment_buffer.emplace_front(raw_text, 0, raw_text.length());
if (parse_special) tokenizer_st_partition(vocab, fragment_buffer);
}
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
switch (vocab.type) {
case LLAMA_VOCAB_TYPE_SPM:
{
BERT tokenizer fixes (#6498) Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
2024-04-09 17:44:08 +00:00
// OG tokenizer behavior:
//
// tokenizer.encode('', add_special_tokens=True) returns [1]
// tokenizer.encode('', add_special_tokens=False) returns []
if (add_special && vocab.special_add_bos != 0) {
GGML_ASSERT(vocab.special_bos_id != -1);
output.push_back(vocab.special_bos_id);
}
tests : multi-thread the tokenizer tests (#5474) * tests : multi-thread the tokenizer tests ggml-ci * unicode : fix data race for unidentified codepoints ggml-ci * unicode : minor style fixes ggml-ci
2024-02-13 13:14:22 +00:00
for (const auto & fragment : fragment_buffer) {
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_RAW_TEXT) {
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
// without adding this leading whitespace, we do not get the same results as the original tokenizer
llama : more tokenizer fixes (#2810) * tests : write a Python tokenizer test (wip) * llama : prefix input text for tokenization with whitespace * llama : distinguish pieces from decoded text + fix detokenization * common : add comments * examples : no longer manually add leading space when tokenizing * tests : use Python to generate tokenizer tests for C++ * tests : add option to tokenize text files ggml-ci * tests : add test-tokenizer-1.py * llama.cpp : fix LF token * hellaswag : move the concat space for clarity * tests : add falcon tests (py + cpp, currently do not pass Unicode) ggml-ci * common : temporary separate llama_detokenize calls for SPM and BPE --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com>
2023-08-27 11:19:19 +00:00
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
// TODO: It's likely possible to get rid of this string copy entirely
// by modifying llm_tokenizer_x to operate with string offsets like pre-tokenizer
// and passing 'add space prefix' as bool argument
//
llama : restore prefix space in llama tokenizer (#4081)
2023-11-15 16:34:47 +00:00
auto raw_text = fragment.raw_text.substr(fragment.offset, fragment.length);
if (&fragment == &fragment_buffer.front()) {
llama : support InternLM2 (#5184) * support InternLM2 inference * add add_space_prefix KV pair
2024-02-01 09:19:51 +00:00
if (vocab.add_space_prefix) {
raw_text = " " + raw_text; // prefix with space if the first token is not special
}
llama : restore prefix space in llama tokenizer (#4081)
2023-11-15 16:34:47 +00:00
}
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
#ifdef PRETOKENIZERDEBUG
llama : fix missing quotes (#4937)
2024-01-14 15:46:00 +00:00
LLAMA_LOG_WARN("TT: (%ld %ld %ld) '%s'\n", raw_text.length(), fragment.offset, fragment.length, raw_text.c_str());
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
#endif
llm_tokenizer_spm tokenizer(vocab);
llama_escape_whitespace(raw_text);
tokenizer.tokenize(raw_text, output);
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
} else { // if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_TOKEN)
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
output.push_back(fragment.token);
}
}
BERT tokenizer fixes (#6498) Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
2024-04-09 17:44:08 +00:00
if (add_special && vocab.special_add_eos == 1) {
GGML_ASSERT(vocab.special_eos_id != -1);
output.push_back(vocab.special_eos_id);
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
} break;
case LLAMA_VOCAB_TYPE_BPE:
{
llama : fix BPE pre-tokenization (#6920) * merged the changes from deepseeker models to main branch * Moved regex patterns to unicode.cpp and updated unicode.h * Moved header files * Resolved issues * added and refactored unicode_regex_split and related functions * Updated/merged the deepseek coder pr * Refactored code * Adding unicode regex mappings * Adding unicode regex function * Added needed functionality, testing remains * Fixed issues * Fixed issue with gpt2 regex custom preprocessor * unicode : fix? unicode_wstring_to_utf8 * lint : fix whitespaces * tests : add tokenizer tests for numbers * unicode : remove redundant headers * tests : remove and rename tokenizer test scripts * tests : add sample usage * gguf-py : reader prints warnings on duplicate keys * llama : towards llama3 tokenization support (wip) * unicode : shot in the dark to fix tests on Windows * unicode : first try custom implementations * convert : add "tokenizer.ggml.pre" GGUF KV (wip) * llama : use new pre-tokenizer type * convert : fix pre-tokenizer type writing * lint : fix * make : add test-tokenizer-0-llama-v3 * wip * models : add llama v3 vocab file * llama : adapt punctuation regex + add llama 3 regex * minor * unicode : set bomb * unicode : set bomb * unicode : always use std::wregex * unicode : support \p{N}, \p{L} and \p{P} natively * unicode : try fix windows * unicode : category support via std::regex * unicode : clean-up * unicode : simplify * convert : add convert-hf-to-gguf-update.py ggml-ci * lint : update * convert : add falcon ggml-ci * unicode : normalize signatures * lint : fix * lint : fix * convert : remove unused functions * convert : add comments * convert : exercise contractions ggml-ci * lint : fix * cmake : refactor test targets * tests : refactor vocab tests ggml-ci * tests : add more vocabs and tests ggml-ci * unicode : cleanup * scripts : ignore new update script in check-requirements.sh * models : add phi-3, mpt, gpt-2, starcoder * tests : disable obsolete ggml-ci * tests : use faster bpe test ggml-ci * llama : more prominent warning for old BPE models * tests : disable test-tokenizer-1-bpe due to slowness ggml-ci --------- Co-authored-by: Jaggzh <jaggz.h@gmail.com> Co-authored-by: Kazim Abrar Mahi <kazimabrarmahi135@gmail.com>
2024-04-29 13:58:41 +00:00
if (add_special && vocab.special_add_bos != 0) {
BERT tokenizer fixes (#6498) Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
2024-04-09 17:44:08 +00:00
GGML_ASSERT(vocab.special_bos_id != -1);
output.push_back(vocab.special_bos_id);
}
tests : multi-thread the tokenizer tests (#5474) * tests : multi-thread the tokenizer tests ggml-ci * unicode : fix data race for unidentified codepoints ggml-ci * unicode : minor style fixes ggml-ci
2024-02-13 13:14:22 +00:00
for (const auto & fragment : fragment_buffer) {
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_RAW_TEXT) {
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
auto raw_text = fragment.raw_text.substr(fragment.offset, fragment.length);
#ifdef PRETOKENIZERDEBUG
llama : fix missing quotes (#4937)
2024-01-14 15:46:00 +00:00
LLAMA_LOG_WARN("TT: (%ld %ld %ld) '%s'\n", raw_text.length(), fragment.offset, fragment.length, raw_text.c_str());
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
#endif
llm_tokenizer_bpe tokenizer(vocab);
tokenizer.tokenize(raw_text, output);
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
} else { // if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_TOKEN)
output.push_back(fragment.token);
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
}
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
}
BERT tokenizer fixes (#6498) Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
2024-04-09 17:44:08 +00:00
GGML_ASSERT(vocab.special_add_eos != 1);
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
} break;
case LLAMA_VOCAB_TYPE_WPM:
{
BERT tokenizer fixes (#6498) Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
2024-04-09 17:44:08 +00:00
if (add_special) {
GGML_ASSERT(vocab.special_cls_id != -1);
output.push_back(vocab.special_cls_id);
}
tests : multi-thread the tokenizer tests (#5474) * tests : multi-thread the tokenizer tests ggml-ci * unicode : fix data race for unidentified codepoints ggml-ci * unicode : minor style fixes ggml-ci
2024-02-13 13:14:22 +00:00
for (const auto & fragment : fragment_buffer) {
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_RAW_TEXT) {
auto raw_text = fragment.raw_text.substr(fragment.offset, fragment.length);
#ifdef PRETOKENIZERDEBUG
LLAMA_LOG_WARN("TT: (%ld %ld %ld) '%s'\n", raw_text.length(), fragment.offset, fragment.length, raw_text.c_str());
#endif
llm_tokenizer_wpm tokenizer(vocab);
tokenizer.tokenize(raw_text, output);
} else { // if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_TOKEN)
tokenizer : special token handling (#3538) * Rewrite special token handling from #1931 * shorten param name, add st verification by type * use offsets instead of copy by substr * formatting, remove copying iterator on delete * llama : normalize code-style * swift fix * print pfx/sfx if verb, main: split pfx input sfx * dont add space when using special tokens * minor : comment + spacing --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-17 15:11:01 +00:00
output.push_back(fragment.token);
}
}
BERT tokenizer fixes (#6498) Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
2024-04-09 17:44:08 +00:00
if (add_special) {
GGML_ASSERT(vocab.special_sep_id != -1);
output.push_back(vocab.special_sep_id);
}
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
} break;
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
case LLAMA_VOCAB_TYPE_NONE:
GGML_ASSERT(false);
build : enable more non-default compiler warnings (#3200)
2023-09-28 21:41:44 +00:00
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
return output;
}
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
//
// grammar - internal
//
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
// Decodes a UTF-8 string which may end in an incomplete sequence. Adds a terminating 0 for use as
// pointer. If an invalid sequence is encountered, returns `llama_partial_utf8.n_remain == -1`.
examples : add GBNF validator program (#5948) * Revising GBNF validator program to be much simpler. * Changing from streams to using cstdio * Adding final newline character.
2024-04-04 07:44:28 +00:00
std::pair<std::vector<uint32_t>, llama_partial_utf8> decode_utf8(
grammar : pre-computed pieces + reserve mem + less string copies (#4330) * reserve space for codepoints * improvement for the appended 0 * used precomputed token text for grammar sample * reserve canidates_decoded * reserve canidates_grammar * remove candidates_decoded * Revert "remove candidates_decoded" This reverts commit 3773328080e6a139ee83198329a13cf4ff61d707. * changed decode_utf8 to take src by ref
2023-12-05 20:55:12 +00:00
const std::string & src,
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
llama_partial_utf8 partial_start) {
static const int lookup[] = { 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 2, 2, 3, 4 };
grammar : pre-computed pieces + reserve mem + less string copies (#4330) * reserve space for codepoints * improvement for the appended 0 * used precomputed token text for grammar sample * reserve canidates_decoded * reserve canidates_grammar * remove candidates_decoded * Revert "remove candidates_decoded" This reverts commit 3773328080e6a139ee83198329a13cf4ff61d707. * changed decode_utf8 to take src by ref
2023-12-05 20:55:12 +00:00
const char * pos = src.c_str();
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
std::vector<uint32_t> code_points;
llama : grammar `reserve` space in `decode_utf8` (#4210) * reserve space for codepoints * improvement for the appended 0
2023-11-25 16:58:23 +00:00
// common english strings have the same number of codepoints and bytes. `+ 1` for the terminating 0.
grammar : pre-computed pieces + reserve mem + less string copies (#4330) * reserve space for codepoints * improvement for the appended 0 * used precomputed token text for grammar sample * reserve canidates_decoded * reserve canidates_grammar * remove candidates_decoded * Revert "remove candidates_decoded" This reverts commit 3773328080e6a139ee83198329a13cf4ff61d707. * changed decode_utf8 to take src by ref
2023-12-05 20:55:12 +00:00
code_points.reserve(src.size() + 1);
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
uint32_t value = partial_start.value;
int n_remain = partial_start.n_remain;
// continue previous decode, if applicable
while (*pos != 0 && n_remain > 0) {
uint8_t next_byte = static_cast<uint8_t>(*pos);
if ((next_byte >> 6) != 2) {
// invalid sequence, abort
code_points.push_back(0);
return std::make_pair(std::move(code_points), llama_partial_utf8{ 0, -1 });
}
value = (value << 6) + (next_byte & 0x3F);
++pos;
--n_remain;
}
if (partial_start.n_remain > 0 && n_remain == 0) {
code_points.push_back(value);
}
// decode any subsequent utf-8 sequences, which may end in an incomplete one
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
while (*pos != 0) {
uint8_t first_byte = static_cast<uint8_t>(*pos);
uint8_t highbits = first_byte >> 4;
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
n_remain = lookup[highbits] - 1;
if (n_remain < 0) {
// invalid sequence, abort
code_points.clear();
code_points.push_back(0);
return std::make_pair(std::move(code_points), llama_partial_utf8{ 0, n_remain });
}
uint8_t mask = (1 << (7 - n_remain)) - 1;
value = first_byte & mask;
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
++pos;
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
while (*pos != 0 && n_remain > 0) {
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
value = (value << 6) + (static_cast<uint8_t>(*pos) & 0x3F);
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
++pos;
--n_remain;
}
if (n_remain == 0) {
code_points.push_back(value);
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
}
}
code_points.push_back(0);
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
return std::make_pair(std::move(code_points), llama_partial_utf8{ value, n_remain });
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
}
// returns true iff pos points to the end of one of the definitions of a rule
static bool llama_grammar_is_end_of_sequence(const llama_grammar_element * pos) {
switch (pos->type) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
case LLAMA_GRETYPE_END: return true; // NOLINT
case LLAMA_GRETYPE_ALT: return true; // NOLINT
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
default: return false;
}
}
// returns true iff chr satisfies the char range at pos (regular or inverse range)
// asserts that pos is pointing to a char range element
static std::pair<bool, const llama_grammar_element *> llama_grammar_match_char(
const llama_grammar_element * pos,
const uint32_t chr) {
bool found = false;
bool is_positive_char = pos->type == LLAMA_GRETYPE_CHAR;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
GGML_ASSERT(is_positive_char || pos->type == LLAMA_GRETYPE_CHAR_NOT); // NOLINT
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
do {
if (pos[1].type == LLAMA_GRETYPE_CHAR_RNG_UPPER) {
// inclusive range, e.g. [a-z]
found = found || (pos->value <= chr && chr <= pos[1].value);
pos += 2;
} else {
// exact char match, e.g. [a] or "a"
found = found || pos->value == chr;
pos += 1;
}
} while (pos->type == LLAMA_GRETYPE_CHAR_ALT);
return std::make_pair(found == is_positive_char, pos);
}
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
// returns true iff some continuation of the given partial UTF-8 sequence could satisfy the char
// range at pos (regular or inverse range)
// asserts that pos is pointing to a char range element
static bool llama_grammar_match_partial_char(
const llama_grammar_element * pos,
const llama_partial_utf8 partial_utf8) {
bool is_positive_char = pos->type == LLAMA_GRETYPE_CHAR;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
GGML_ASSERT(is_positive_char || pos->type == LLAMA_GRETYPE_CHAR_NOT);
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
uint32_t partial_value = partial_utf8.value;
int n_remain = partial_utf8.n_remain;
// invalid sequence or 7-bit char split across 2 bytes (overlong)
if (n_remain < 0 || (n_remain == 1 && partial_value < 2)) {
return false;
}
// range of possible code points this partial UTF-8 sequence could complete to
uint32_t low = partial_value << (n_remain * 6);
uint32_t high = low | ((1 << (n_remain * 6)) - 1);
if (low == 0) {
if (n_remain == 2) {
low = 1 << 11;
} else if (n_remain == 3) {
low = 1 << 16;
}
}
do {
if (pos[1].type == LLAMA_GRETYPE_CHAR_RNG_UPPER) {
// inclusive range, e.g. [a-z]
if (pos->value <= high && low <= pos[1].value) {
return is_positive_char;
}
pos += 2;
} else {
// exact char match, e.g. [a] or "a"
if (low <= pos->value && pos->value <= high) {
return is_positive_char;
}
pos += 1;
}
} while (pos->type == LLAMA_GRETYPE_CHAR_ALT);
return !is_positive_char;
}
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
// transforms a grammar pushdown stack into N possible stacks, all ending
// at a character range (terminal element)
static void llama_grammar_advance_stack(
const std::vector<std::vector<llama_grammar_element>> & rules,
const std::vector<const llama_grammar_element *> & stack,
std::vector<std::vector<const llama_grammar_element *>> & new_stacks) {
if (stack.empty()) {
Optimization: eliminate addition of redundant stacks when advancing grammar. (#6616)
2024-04-12 01:44:50 +00:00
if (std::find(new_stacks.begin(), new_stacks.end(), stack) == new_stacks.end()) {
new_stacks.emplace_back(stack);
}
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
return;
}
const llama_grammar_element * pos = stack.back();
switch (pos->type) {
case LLAMA_GRETYPE_RULE_REF: {
const size_t rule_id = static_cast<size_t>(pos->value);
const llama_grammar_element * subpos = rules[rule_id].data();
do {
// init new stack without the top (pos)
std::vector<const llama_grammar_element *> new_stack(stack.begin(), stack.end() - 1);
if (!llama_grammar_is_end_of_sequence(pos + 1)) {
// if this rule ref is followed by another element, add that to stack
new_stack.push_back(pos + 1);
}
if (!llama_grammar_is_end_of_sequence(subpos)) {
// if alternate is nonempty, add to stack
new_stack.push_back(subpos);
}
llama_grammar_advance_stack(rules, new_stack, new_stacks);
while (!llama_grammar_is_end_of_sequence(subpos)) {
// scan to end of alternate def
subpos++;
}
if (subpos->type == LLAMA_GRETYPE_ALT) {
// there's another alternate def of this rule to process
subpos++;
} else {
break;
}
} while (true);
break;
}
case LLAMA_GRETYPE_CHAR:
case LLAMA_GRETYPE_CHAR_NOT:
Optimization: eliminate addition of redundant stacks when advancing grammar. (#6616)
2024-04-12 01:44:50 +00:00
if (std::find(new_stacks.begin(), new_stacks.end(), stack) == new_stacks.end()) {
// only add the stack if it's not a duplicate of one we already have
new_stacks.emplace_back(stack);
}
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
break;
default:
// end of alternate (LLAMA_GRETYPE_END, LLAMA_GRETYPE_ALT) or middle of char range
// (LLAMA_GRETYPE_CHAR_ALT, LLAMA_GRETYPE_CHAR_RNG_UPPER); stack should never be left on
// those
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
GGML_ASSERT(false);
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
}
}
// takes a set of possible pushdown stacks on a grammar, which are required to
// be positioned at a character range (see `llama_grammar_advance_stack`), and
// produces the N possible stacks if the given char is accepted at those
// positions
grammars: 1.5x faster inference w/ complex grammars (vector reserves / reuses) (#6609) * grammars: reserve rejects & next candidates * grammars: reuse new_stacks * grammars: fix missing sig change in llama.h * grammars: fix test (api changed) * grammars: update gbnf-validator.cpp * grammars: simpler syntax (no swap)
2024-04-11 18:47:34 +00:00
void llama_grammar_accept(
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
const std::vector<std::vector<llama_grammar_element>> & rules,
const std::vector<std::vector<const llama_grammar_element *>> & stacks,
grammars: 1.5x faster inference w/ complex grammars (vector reserves / reuses) (#6609) * grammars: reserve rejects & next candidates * grammars: reuse new_stacks * grammars: fix missing sig change in llama.h * grammars: fix test (api changed) * grammars: update gbnf-validator.cpp * grammars: simpler syntax (no swap)
2024-04-11 18:47:34 +00:00
const uint32_t chr,
std::vector<std::vector<const llama_grammar_element *>> & new_stacks) {
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
grammars: 1.5x faster inference w/ complex grammars (vector reserves / reuses) (#6609) * grammars: reserve rejects & next candidates * grammars: reuse new_stacks * grammars: fix missing sig change in llama.h * grammars: fix test (api changed) * grammars: update gbnf-validator.cpp * grammars: simpler syntax (no swap)
2024-04-11 18:47:34 +00:00
new_stacks.clear();
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
for (const auto & stack : stacks) {
if (stack.empty()) {
continue;
}
auto match = llama_grammar_match_char(stack.back(), chr);
if (match.first) {
const llama_grammar_element * pos = match.second;
// update top of stack to next element, if any
std::vector<const llama_grammar_element *> new_stack(stack.begin(), stack.end() - 1);
if (!llama_grammar_is_end_of_sequence(pos)) {
new_stack.push_back(pos);
}
llama_grammar_advance_stack(rules, new_stack, new_stacks);
}
}
}
static std::vector<llama_grammar_candidate> llama_grammar_reject_candidates(
const std::vector<std::vector<llama_grammar_element>> & rules,
const std::vector<std::vector<const llama_grammar_element *>> & stacks,
const std::vector<llama_grammar_candidate> & candidates);
static std::vector<llama_grammar_candidate> llama_grammar_reject_candidates_for_stack(
const std::vector<std::vector<llama_grammar_element>> & rules,
const std::vector<const llama_grammar_element *> & stack,
const std::vector<llama_grammar_candidate> & candidates) {
std::vector<llama_grammar_candidate> rejects;
grammars: 1.5x faster inference w/ complex grammars (vector reserves / reuses) (#6609) * grammars: reserve rejects & next candidates * grammars: reuse new_stacks * grammars: fix missing sig change in llama.h * grammars: fix test (api changed) * grammars: update gbnf-validator.cpp * grammars: simpler syntax (no swap)
2024-04-11 18:47:34 +00:00
rejects.reserve(candidates.size());
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
if (stack.empty()) {
ggml : fix rope + llama minor optimizations (#3560) * Minor fixes and fixed memleak * Using const auto references in range-based loop C++17
2023-10-20 10:02:12 +00:00
for (const auto & tok : candidates) {
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
if (*tok.code_points != 0 || tok.partial_utf8.n_remain != 0) {
rejects.push_back(tok);
}
}
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
return rejects;
}
const llama_grammar_element * stack_pos = stack.back();
std::vector<llama_grammar_candidate> next_candidates;
grammars: 1.5x faster inference w/ complex grammars (vector reserves / reuses) (#6609) * grammars: reserve rejects & next candidates * grammars: reuse new_stacks * grammars: fix missing sig change in llama.h * grammars: fix test (api changed) * grammars: update gbnf-validator.cpp * grammars: simpler syntax (no swap)
2024-04-11 18:47:34 +00:00
next_candidates.reserve(candidates.size());
ggml : fix rope + llama minor optimizations (#3560) * Minor fixes and fixed memleak * Using const auto references in range-based loop C++17
2023-10-20 10:02:12 +00:00
for (const auto & tok : candidates) {
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
if (*tok.code_points == 0) {
// reached end of full codepoints in token, reject iff it ended in a partial sequence
// that cannot satisfy this position in grammar
if (tok.partial_utf8.n_remain != 0 &&
!llama_grammar_match_partial_char(stack_pos, tok.partial_utf8)) {
rejects.push_back(tok);
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
}
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
} else if (llama_grammar_match_char(stack_pos, *tok.code_points).first) {
next_candidates.push_back({ tok.index, tok.code_points + 1, tok.partial_utf8 });
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
} else {
rejects.push_back(tok);
}
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
const auto * stack_pos_after = llama_grammar_match_char(stack_pos, 0).second;
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
// update top of stack to next element, if any
std::vector<const llama_grammar_element *> stack_after(stack.begin(), stack.end() - 1);
if (!llama_grammar_is_end_of_sequence(stack_pos_after)) {
stack_after.push_back(stack_pos_after);
}
std::vector<std::vector<const llama_grammar_element *>> next_stacks;
llama_grammar_advance_stack(rules, stack_after, next_stacks);
auto next_rejects = llama_grammar_reject_candidates(rules, next_stacks, next_candidates);
ggml : fix rope + llama minor optimizations (#3560) * Minor fixes and fixed memleak * Using const auto references in range-based loop C++17
2023-10-20 10:02:12 +00:00
for (const auto & tok : next_rejects) {
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
rejects.push_back({ tok.index, tok.code_points - 1, tok.partial_utf8 });
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
}
return rejects;
}
static std::vector<llama_grammar_candidate> llama_grammar_reject_candidates(
const std::vector<std::vector<llama_grammar_element>> & rules,
const std::vector<std::vector<const llama_grammar_element *>> & stacks,
const std::vector<llama_grammar_candidate> & candidates) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
GGML_ASSERT(!stacks.empty()); // REVIEW
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
if (candidates.empty()) {
return std::vector<llama_grammar_candidate>();
}
auto rejects = llama_grammar_reject_candidates_for_stack(rules, stacks.front(), candidates);
for (size_t i = 1, size = stacks.size(); i < size; ++i) {
rejects = llama_grammar_reject_candidates_for_stack(rules, stacks[i], rejects);
}
return rejects;
}
//
// grammar - external
//
struct llama_grammar * llama_grammar_init(
const llama_grammar_element ** rules,
size_t n_rules,
size_t start_rule_index) {
const llama_grammar_element * pos;
// copy rule definitions into vectors
std::vector<std::vector<llama_grammar_element>> vec_rules(n_rules);
for (size_t i = 0; i < n_rules; i++) {
for (pos = rules[i]; pos->type != LLAMA_GRETYPE_END; pos++) {
vec_rules[i].push_back(*pos);
}
vec_rules[i].push_back({LLAMA_GRETYPE_END, 0});
}
// loop over alternates of start rule to build initial stacks
std::vector<std::vector<const llama_grammar_element *>> stacks;
grammar : fix unnecessarily retained pointer to rules (#6003)
2024-03-11 19:59:03 +00:00
pos = vec_rules[start_rule_index].data();
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
do {
std::vector<const llama_grammar_element *> stack;
if (!llama_grammar_is_end_of_sequence(pos)) {
// if alternate is nonempty, add to stack
stack.push_back(pos);
}
llama_grammar_advance_stack(vec_rules, stack, stacks);
while (!llama_grammar_is_end_of_sequence(pos)) {
// scan to end of alternate def
pos++;
}
if (pos->type == LLAMA_GRETYPE_ALT) {
// there's another alternate def of this rule to process
pos++;
} else {
break;
}
} while (true);
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
return new llama_grammar{ std::move(vec_rules), std::move(stacks), {} };
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
}
void llama_grammar_free(struct llama_grammar * grammar) {
delete grammar;
}
speculative : add grammar support (#2991) * speculative : add grammar support * grammars : add json_arr.gbnf * grammar : add comments to new grammar file * grammar : remove one nested level * common : warm-up with 2 tokens - seems to work better * speculative : print draft token pieces * speculative : reuse grammar parser + better logs and comments * speculative : avoid grammar_mem * make : fix speculative build
2023-09-05 05:46:17 +00:00
struct llama_grammar * llama_grammar_copy(const struct llama_grammar * grammar) {
llama_grammar * result = new llama_grammar{ grammar->rules, grammar->stacks, grammar->partial_utf8 };
// redirect elements in stacks to point to new rules
for (size_t is = 0; is < result->stacks.size(); is++) {
for (size_t ie = 0; ie < result->stacks[is].size(); ie++) {
for (size_t ir0 = 0; ir0 < grammar->rules.size(); ir0++) {
for (size_t ir1 = 0; ir1 < grammar->rules[ir0].size(); ir1++) {
if (grammar->stacks[is][ie] == &grammar->rules[ir0][ir1]) {
result->stacks[is][ie] = &result->rules[ir0][ir1];
}
}
}
}
}
return result;
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
//
// sampling
//
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
void llama_set_rng_seed(struct llama_context * ctx, uint32_t seed) {
if (seed == LLAMA_DEFAULT_SEED) {
seed = time(NULL);
}
ctx->rng.seed(seed);
}
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
void llama_sample_softmax(struct llama_context * ctx, llama_token_data_array * candidates) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
GGML_ASSERT(candidates->size > 0);
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
const int64_t t_start_sample_us = ggml_time_us();
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
// Sort the logits in descending order
if (!candidates->sorted) {
std::sort(candidates->data, candidates->data + candidates->size, [](const llama_token_data & a, const llama_token_data & b) {
return a.logit > b.logit;
});
candidates->sorted = true;
}
float max_l = candidates->data[0].logit;
float cum_sum = 0.0f;
for (size_t i = 0; i < candidates->size; ++i) {
float p = expf(candidates->data[i].logit - max_l);
candidates->data[i].p = p;
cum_sum += p;
}
for (size_t i = 0; i < candidates->size; ++i) {
candidates->data[i].p /= cum_sum;
}
if (ctx) {
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
void llama_sample_top_k(struct llama_context * ctx, llama_token_data_array * candidates, int32_t k, size_t min_keep) {
Tests for min_p, sampling queue (#5147)
2024-01-28 08:35:14 +00:00
// TODO: move bucket sort to separate function so that top_p/tail_free/typical/softmax first is equally fast
// if (k >= (int32_t)candidates->size) {
// return;
// }
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
const int64_t t_start_sample_us = ggml_time_us();
fix trailing whitespace (#5407)
2024-02-08 10:36:54 +00:00
sampling: fix top_k <= 0 (#5388) * sampling: fix top_k <= 0 * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-08 08:46:30 +00:00
if (k <= 0) {
k = candidates->size;
}
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
k = std::max(k, (int) min_keep);
k = std::min(k, (int) candidates->size);
// Sort scores in descending order
if (!candidates->sorted) {
auto comp = [](const llama_token_data & a, const llama_token_data & b) {
return a.logit > b.logit;
};
Another bucket sort (#5109) * Initial bucket sort * Bucket sort: slightly better version * Bucket sort: another minor improvement --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-26 07:14:39 +00:00
if (k <= 128) {
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
std::partial_sort(candidates->data, candidates->data + k, candidates->data + candidates->size, comp);
Another bucket sort (#5109) * Initial bucket sort * Bucket sort: slightly better version * Bucket sort: another minor improvement --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-26 07:14:39 +00:00
} else {
constexpr int nbuckets = 128;
constexpr float bucket_low = -10.0f;
constexpr float bucket_high = 10.0f;
constexpr float bucket_scale = nbuckets/(bucket_high - bucket_low);
constexpr float bucker_inter = -bucket_low * bucket_scale;
std::vector<int> bucket_idx(candidates->size);
std::vector<int> histo(nbuckets, 0);
for (int i = 0; i < (int)candidates->size; ++i) {
const float val = candidates->data[i].logit;
int ib = int(bucket_scale * val + bucker_inter); //nbuckets * (val - bucket_low) / (bucket_high - bucket_low);
ib = std::max(0, std::min(nbuckets-1, ib));
bucket_idx[i] = ib;
++histo[ib];
}
int nhave = 0;
int ib = nbuckets - 1;
for ( ; ib >= 0; --ib) {
nhave += histo[ib];
if (nhave >= k) break;
}
std::vector<llama_token_data> tmp_tokens(nhave);
auto ptr = tmp_tokens.data();
std::vector<llama_token_data*> bucket_ptrs;
bucket_ptrs.reserve(nbuckets - ib);
for (int j = nbuckets - 1; j >= ib; --j) {
bucket_ptrs.push_back(ptr);
ptr += histo[j];
}
for (int i = 0; i < (int)candidates->size; ++i) {
int j = bucket_idx[i];
if (j >= ib) {
*bucket_ptrs[nbuckets-1-j]++ = candidates->data[i];
}
}
ptr = tmp_tokens.data();
int ndone = 0;
for (int j = nbuckets-1; j > ib; --j) {
std::sort(ptr, ptr + histo[j], comp);
ptr += histo[j];
ndone += histo[j];
}
std::partial_sort(ptr, ptr + k - ndone, ptr + histo[ib], comp);
std::memcpy(candidates->data, tmp_tokens.data(), k*sizeof(llama_token_data));
Define non-positive temperature behavior (#720)
2023-04-03 00:19:04 +00:00
}
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
candidates->sorted = true;
Define non-positive temperature behavior (#720)
2023-04-03 00:19:04 +00:00
}
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
candidates->size = k;
Define non-positive temperature behavior (#720)
2023-04-03 00:19:04 +00:00
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
if (ctx) {
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
}
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
void llama_sample_top_p(struct llama_context * ctx, llama_token_data_array * candidates, float p, size_t min_keep) {
if (p >= 1.0f) {
return;
}
llama_sample_softmax(ctx, candidates);
llama: Don't double count the sampling time (#2107)
2023-07-05 10:31:23 +00:00
const int64_t t_start_sample_us = ggml_time_us();
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
// Compute the cumulative probabilities
float cum_sum = 0.0f;
size_t last_idx = candidates->size;
for (size_t i = 0; i < candidates->size; ++i) {
cum_sum += candidates->data[i].p;
llama : fix top-p sampling to match the canonical definition (#1953) * Fix top-p sampling to match the standard definition (smallest set that has probability mass at least p, not largest set with probability mass less than p) * top-p: correct gt to gte * add test for correct top-p behavior
2023-06-24 10:15:01 +00:00
// Check if the running sum is at least p or if we have kept at least min_keep tokens
// we set the last index to i+1 to indicate that the current iterate should be included in the set
if (cum_sum >= p && i + 1 >= min_keep) {
last_idx = i + 1;
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
break;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
}
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
// Resize the output vector to keep only the top-p tokens
candidates->size = last_idx;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
if (ctx) {
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
}
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
samplers : Min-P sampler implementation [alternative to Top P/Top K] (#3841) * Introduce the new Min-P sampler by @kalomaze The Min-P sampling method was designed as an alternative to Top-P, and aims to ensure a balance of quality and variety. The parameter *p* represents the minimum probability for a token to be considered, relative to the probability of the most likely token. * Min-P enabled and set to 0.05 default --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: cebtenzzre <cebtenzzre@gmail.com>
2023-10-31 19:44:49 +00:00
void llama_sample_min_p(struct llama_context * ctx, llama_token_data_array * candidates, float p, size_t min_keep) {
if (p <= 0.0f || !candidates->size) {
return;
}
const int64_t t_start_sample_us = ggml_time_us();
Apply min_p to unsorted tokens (#5115)
2024-01-28 08:59:49 +00:00
bool min_p_applied = false;
// if the candidates aren't sorted, try the unsorted implementation first
if (!candidates->sorted) {
std::vector<llama_token_data> filtered_tokens;
float max_logit = -FLT_MAX;
for (size_t i = 0; i < candidates->size; ++i) {
max_logit = std::max(max_logit, candidates->data[i].logit);
}
const float min_logit = max_logit + logf(p); // min logit for p_i >= p * p_max
for (size_t i = 0; i < candidates->size; ++i) {
if (candidates->data[i].logit >= min_logit) {
filtered_tokens.push_back(candidates->data[i]);
}
}
samplers : Min-P sampler implementation [alternative to Top P/Top K] (#3841) * Introduce the new Min-P sampler by @kalomaze The Min-P sampling method was designed as an alternative to Top-P, and aims to ensure a balance of quality and variety. The parameter *p* represents the minimum probability for a token to be considered, relative to the probability of the most likely token. * Min-P enabled and set to 0.05 default --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: cebtenzzre <cebtenzzre@gmail.com>
2023-10-31 19:44:49 +00:00
Apply min_p to unsorted tokens (#5115)
2024-01-28 08:59:49 +00:00
// if we have enough values the operation was a success
if (filtered_tokens.size() >= min_keep) {
memcpy(candidates->data, filtered_tokens.data(), filtered_tokens.size()*sizeof(llama_token_data));
candidates->size = filtered_tokens.size();
min_p_applied = true;
samplers : Min-P sampler implementation [alternative to Top P/Top K] (#3841) * Introduce the new Min-P sampler by @kalomaze The Min-P sampling method was designed as an alternative to Top-P, and aims to ensure a balance of quality and variety. The parameter *p* represents the minimum probability for a token to be considered, relative to the probability of the most likely token. * Min-P enabled and set to 0.05 default --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: cebtenzzre <cebtenzzre@gmail.com>
2023-10-31 19:44:49 +00:00
}
}
Apply min_p to unsorted tokens (#5115)
2024-01-28 08:59:49 +00:00
// if the candidates are sorted or the unsorted implementation failed, use this implementation
if (!min_p_applied) {
// Sort the logits in descending order
if (!candidates->sorted) {
std::sort(candidates->data, candidates->data + candidates->size, [](const llama_token_data & a, const llama_token_data & b) {
return a.logit > b.logit;
});
candidates->sorted = true;
}
const float min_logit = candidates->data[0].logit + logf(p); // min logit for p_i >= p * p_max
size_t i = 1; // first token always matches
for (; i < candidates->size; ++i) {
if (candidates->data[i].logit < min_logit && i >= min_keep) {
break; // prob too small
}
}
// Resize the output vector to keep only the matching tokens
candidates->size = i;
}
samplers : Min-P sampler implementation [alternative to Top P/Top K] (#3841) * Introduce the new Min-P sampler by @kalomaze The Min-P sampling method was designed as an alternative to Top-P, and aims to ensure a balance of quality and variety. The parameter *p* represents the minimum probability for a token to be considered, relative to the probability of the most likely token. * Min-P enabled and set to 0.05 default --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: cebtenzzre <cebtenzzre@gmail.com>
2023-10-31 19:44:49 +00:00
if (ctx) {
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
}
}
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
void llama_sample_tail_free(struct llama_context * ctx, llama_token_data_array * candidates, float z, size_t min_keep) {
if (z >= 1.0f || candidates->size <= 2) {
return;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
llama_sample_softmax(nullptr, candidates);
llama: Don't double count the sampling time (#2107)
2023-07-05 10:31:23 +00:00
const int64_t t_start_sample_us = ggml_time_us();
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
// Compute the first and second derivatives
std::vector<float> first_derivatives(candidates->size - 1);
std::vector<float> second_derivatives(candidates->size - 2);
for (size_t i = 0; i < first_derivatives.size(); ++i) {
first_derivatives[i] = candidates->data[i].p - candidates->data[i + 1].p;
}
for (size_t i = 0; i < second_derivatives.size(); ++i) {
second_derivatives[i] = first_derivatives[i] - first_derivatives[i + 1];
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
// Calculate absolute value of second derivatives
for (size_t i = 0; i < second_derivatives.size(); ++i) {
llama : use std::abs in llama_sample_tail_free (#2800) Plain 'abs' casts the input to int.
2023-08-26 16:53:52 +00:00
second_derivatives[i] = std::abs(second_derivatives[i]);
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
}
// Normalize the second derivatives
ci : integrate with ggml-org/ci (#2250) * ci : run ctest ggml-ci * ci : add open llama 3B-v2 tests ggml-ci * ci : disable wget progress output ggml-ci * ci : add open llama 3B-v2 tg tests for q4 and q5 quantizations ggml-ci * tests : try to fix tail free sampling test ggml-ci * ci : add K-quants ggml-ci * ci : add short perplexity tests ggml-ci * ci : add README.md * ppl : add --chunks argument to limit max number of chunks ggml-ci * ci : update README
2023-07-18 11:24:43 +00:00
{
const float second_derivatives_sum = std::accumulate(second_derivatives.begin(), second_derivatives.end(), 0.0f);
if (second_derivatives_sum > 1e-6f) {
for (float & value : second_derivatives) {
value /= second_derivatives_sum;
}
} else {
for (float & value : second_derivatives) {
value = 1.0f / second_derivatives.size();
}
}
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
}
float cum_sum = 0.0f;
size_t last_idx = candidates->size;
for (size_t i = 0; i < second_derivatives.size(); ++i) {
cum_sum += second_derivatives[i];
// Check if the running sum is greater than z or if we have kept at least min_keep tokens
if (cum_sum > z && i >= min_keep) {
last_idx = i;
break;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
}
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
// Resize the output vector to keep only the tokens above the tail location
candidates->size = last_idx;
if (ctx) {
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
}
}
void llama_sample_typical(struct llama_context * ctx, llama_token_data_array * candidates, float p, size_t min_keep) {
// Reference implementation:
// https://github.com/huggingface/transformers/compare/main...cimeister:typical-sampling:typical-pr
if (p >= 1.0f) {
return;
}
// Compute the softmax of logits and calculate entropy
llama_sample_softmax(nullptr, candidates);
llama: Don't double count the sampling time (#2107)
2023-07-05 10:31:23 +00:00
const int64_t t_start_sample_us = ggml_time_us();
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
float entropy = 0.0f;
for (size_t i = 0; i < candidates->size; ++i) {
entropy += -candidates->data[i].p * logf(candidates->data[i].p);
}
// Compute the absolute difference between negative log probability and entropy for each candidate
std::vector<float> shifted_scores;
for (size_t i = 0; i < candidates->size; ++i) {
float shifted_score = fabsf(-logf(candidates->data[i].p) - entropy);
shifted_scores.push_back(shifted_score);
}
// Sort tokens based on the shifted_scores and their corresponding indices
std::vector<size_t> indices(candidates->size);
std::iota(indices.begin(), indices.end(), 0);
std::sort(indices.begin(), indices.end(), [&](size_t a, size_t b) {
return shifted_scores[a] < shifted_scores[b];
});
// Compute the cumulative probabilities
float cum_sum = 0.0f;
size_t last_idx = indices.size();
for (size_t i = 0; i < indices.size(); ++i) {
size_t idx = indices[i];
cum_sum += candidates->data[idx].p;
// Check if the running sum is greater than typical or if we have kept at least min_keep tokens
if (cum_sum > p && i >= min_keep - 1) {
last_idx = i + 1;
break;
}
}
// Resize the output vector to keep only the locally typical tokens
std::vector<llama_token_data> new_candidates;
for (size_t i = 0; i < last_idx; ++i) {
size_t idx = indices[i];
new_candidates.push_back(candidates->data[idx]);
}
// Replace the data in candidates with the new_candidates data
std::copy(new_candidates.begin(), new_candidates.end(), candidates->data);
candidates->size = new_candidates.size();
llama : fix typical sampling (#4261) Typical sampling was broken because after copying new_candidates into canditates, the "sorted" bool is left at "true", but the new data is no longer sorted according to probability. Patch to set "sorted" to false. Test: Generating with temp=0.0001 (approx. argmax) should generate the same sequence at typical>=1.0 and typical=0.9999 (approx. disabled, but enters the typical sampling codepath).
2023-11-30 21:40:23 +00:00
candidates->sorted = false;
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
if (ctx) {
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
}
}
llama : dynamic temperature sampling (#4972) * implemented dynamic temperature sampling from koboldcpp * removed trailing whitespace * removed unused temp parameter in llama_sample_entropy * exposed exponent_val in dynamic temp sampler * added debug check for printf statements * use nullptr in llama_sample_softmax call during llama_sample_entropy this avoids counting the time taken stats twice Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * return earlier if there is only 1 candiate (i.e. max_entropy == 0) * reformat 't' case in llama_sample_queue Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * check for one or zero candidates case in llama_sample_entropy --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2024-01-25 20:06:22 +00:00
void llama_sample_entropy(struct llama_context * ctx, llama_token_data_array * candidates_p, float min_temp, float max_temp, float exponent_val) {
const int64_t t_start_sample_us = ggml_time_us();
// no need to do anything if there is only one (or zero) candidates
if(candidates_p->size <= 1) {
return;
}
// Calculate maximum possible entropy
float max_entropy = -logf(1.0f / candidates_p->size);
llama_sample_softmax(nullptr, candidates_p);
// Calculate entropy of the softmax probabilities
float entropy = 0.0f;
for (size_t i = 0; i < candidates_p->size; ++i) {
float prob = candidates_p->data[i].p;
if (prob > 0.0f) { // Ensure no log(0)
entropy -= prob * logf(prob);
}
}
// Normalize the entropy (max_entropy cannot be 0 here because we checked candidates_p->size != 1 above)
float normalized_entropy = entropy / max_entropy;
// Map the normalized entropy to the desired temperature range using the power function
float dyn_temp = min_temp + (max_temp - min_temp) * powf(normalized_entropy, exponent_val);
#ifdef DEBUG
LLAMA_LOG_INFO("Your text maxtemp value is: %f\n", max_temp);
LLAMA_LOG_INFO("Entropy: %f\n", entropy);
LLAMA_LOG_INFO("Max Possible Entropy: %f\n", max_entropy);
LLAMA_LOG_INFO("Normalized Entropy: %f\n", normalized_entropy);
LLAMA_LOG_INFO("Exponent: %f\n", exponent_val);
LLAMA_LOG_INFO("Dynamic Temperature (dyn_temp): %f\n", dyn_temp);
#endif
// Apply the dynamically calculated temperature scaling
for (size_t i = 0; i < candidates_p->size; ++i) {
candidates_p->data[i].logit /= dyn_temp;
}
// Re-compute softmax probabilities after scaling logits with dynamic temperature
double max_l_double = candidates_p->data[0].logit;
double cum_sum_double = 0.0;
for (size_t i = 0; i < candidates_p->size; ++i) {
double p = exp(candidates_p->data[i].logit - max_l_double);
candidates_p->data[i].p = p; // Store the scaled probability
cum_sum_double += p;
}
for (size_t i = 0; i < candidates_p->size; ++i) {
candidates_p->data[i].p /= cum_sum_double; // Re-normalize the probabilities
}
#ifdef DEBUG
// Print the updated top 25 probabilities after temperature scaling
LLAMA_LOG_INFO("\nUpdated Top 25 Probabilities After Dynamic Temperature Scaling (in percentages):\n");
for (size_t i = 0; i < 25 && i < candidates_p->size; ++i) {
LLAMA_LOG_INFO("Token %zu: %f%%\n", i + 1, candidates_p->data[i].p * 100.0f);
}
#endif
if (ctx) {
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
}
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
void llama_sample_temp(struct llama_context * ctx, llama_token_data_array * candidates_p, float temp) {
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
const int64_t t_start_sample_us = ggml_time_us();
for (size_t i = 0; i < candidates_p->size; ++i) {
candidates_p->data[i].logit /= temp;
}
if (ctx) {
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
}
}
sampling : refactor init to use llama_sampling_params (#3696) * sampling : refactor init to use llama_sampling_params * llama : combine repetition, frequency and presence penalties in 1 call * examples : remove embd-input and gptneox-wip * sampling : rename penalty params + reduce size of "prev" vector * sampling : add llama_sampling_print helper * sampling : hide prev behind API and apply #3661 ggml-ci
2023-10-20 18:07:23 +00:00
void llama_sample_repetition_penalties(
struct llama_context * ctx,
llama_token_data_array * candidates,
const llama_token * last_tokens,
size_t penalty_last_n,
float penalty_repeat,
float penalty_freq,
float penalty_present) {
if (penalty_last_n == 0 || (penalty_repeat == 1.0f && penalty_freq == 0.0f && penalty_present == 0.0f)) {
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
return;
}
const int64_t t_start_sample_us = ggml_time_us();
// Create a frequency map to count occurrences of each token in last_tokens
std::unordered_map<llama_token, int> token_count;
sampling : refactor init to use llama_sampling_params (#3696) * sampling : refactor init to use llama_sampling_params * llama : combine repetition, frequency and presence penalties in 1 call * examples : remove embd-input and gptneox-wip * sampling : rename penalty params + reduce size of "prev" vector * sampling : add llama_sampling_print helper * sampling : hide prev behind API and apply #3661 ggml-ci
2023-10-20 18:07:23 +00:00
for (size_t i = 0; i < penalty_last_n; ++i) {
token_count[last_tokens[i]]++;
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
}
// Apply frequency and presence penalties to the candidates
for (size_t i = 0; i < candidates->size; ++i) {
sampling : refactor init to use llama_sampling_params (#3696) * sampling : refactor init to use llama_sampling_params * llama : combine repetition, frequency and presence penalties in 1 call * examples : remove embd-input and gptneox-wip * sampling : rename penalty params + reduce size of "prev" vector * sampling : add llama_sampling_print helper * sampling : hide prev behind API and apply #3661 ggml-ci
2023-10-20 18:07:23 +00:00
const auto token_iter = token_count.find(candidates->data[i].id);
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
if (token_iter == token_count.end()) {
continue;
}
sampling : refactor init to use llama_sampling_params (#3696) * sampling : refactor init to use llama_sampling_params * llama : combine repetition, frequency and presence penalties in 1 call * examples : remove embd-input and gptneox-wip * sampling : rename penalty params + reduce size of "prev" vector * sampling : add llama_sampling_print helper * sampling : hide prev behind API and apply #3661 ggml-ci
2023-10-20 18:07:23 +00:00
const int count = token_iter->second;
// The academic publication that described this technique actually just only divided, but that would cause tokens with negative logits to become more likely, which is obviously wrong.
// This is common fix for this problem, which is to multiply by the penalty instead of dividing.
if (candidates->data[i].logit <= 0) {
candidates->data[i].logit *= penalty_repeat;
} else {
candidates->data[i].logit /= penalty_repeat;
}
candidates->data[i].logit -= float(count) * penalty_freq + float(count > 0) * penalty_present;
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
}
candidates->sorted = false;
if (ctx) {
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
}
}
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
void llama_sample_grammar(struct llama_context * ctx, llama_token_data_array * candidates, const struct llama_grammar * grammar) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
GGML_ASSERT(ctx);
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
const int64_t t_start_sample_us = ggml_time_us();
llama : support Llama 3 HF conversion (#6745) * Support Llama 3 conversion The tokenizer is BPE. * style * Accept suggestion Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> * llama : add llama_token_is_eog() ggml-ci * llama : auto-detect more EOT tokens when missing in KV data * convert : replacing EOS token is a hack * llama : fix codegemma EOT token + add TODOs * llama : fix model type string for 8B model --------- Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-21 11:50:41 +00:00
bool allow_eog = false;
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
for (const auto & stack : grammar->stacks) {
if (stack.empty()) {
llama : support Llama 3 HF conversion (#6745) * Support Llama 3 conversion The tokenizer is BPE. * style * Accept suggestion Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> * llama : add llama_token_is_eog() ggml-ci * llama : auto-detect more EOT tokens when missing in KV data * convert : replacing EOS token is a hack * llama : fix codegemma EOT token + add TODOs * llama : fix model type string for 8B model --------- Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-21 11:50:41 +00:00
allow_eog = true;
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
break;
}
}
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
std::vector<std::pair<std::vector<uint32_t>, llama_partial_utf8>> candidates_decoded;
grammar : pre-computed pieces + reserve mem + less string copies (#4330) * reserve space for codepoints * improvement for the appended 0 * used precomputed token text for grammar sample * reserve canidates_decoded * reserve canidates_grammar * remove candidates_decoded * Revert "remove candidates_decoded" This reverts commit 3773328080e6a139ee83198329a13cf4ff61d707. * changed decode_utf8 to take src by ref
2023-12-05 20:55:12 +00:00
candidates_decoded.reserve(candidates->size);
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
std::vector<llama_grammar_candidate> candidates_grammar;
grammar : pre-computed pieces + reserve mem + less string copies (#4330) * reserve space for codepoints * improvement for the appended 0 * used precomputed token text for grammar sample * reserve canidates_decoded * reserve canidates_grammar * remove candidates_decoded * Revert "remove candidates_decoded" This reverts commit 3773328080e6a139ee83198329a13cf4ff61d707. * changed decode_utf8 to take src by ref
2023-12-05 20:55:12 +00:00
candidates_grammar.reserve(candidates->size);
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
for (size_t i = 0; i < candidates->size; ++i) {
llama : more tokenizer fixes (#2810) * tests : write a Python tokenizer test (wip) * llama : prefix input text for tokenization with whitespace * llama : distinguish pieces from decoded text + fix detokenization * common : add comments * examples : no longer manually add leading space when tokenizing * tests : use Python to generate tokenizer tests for C++ * tests : add option to tokenize text files ggml-ci * tests : add test-tokenizer-1.py * llama.cpp : fix LF token * hellaswag : move the concat space for clarity * tests : add falcon tests (py + cpp, currently do not pass Unicode) ggml-ci * common : temporary separate llama_detokenize calls for SPM and BPE --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com>
2023-08-27 11:19:19 +00:00
const llama_token id = candidates->data[i].id;
llama : add option to render special/control tokens (#6807) * make : fix common dep on llama.h * llama : add option to render special tokens * readme : add API change notice ggml-ci * swift : fix build
2024-04-21 15:36:45 +00:00
const std::string piece = llama_token_to_piece(ctx, id, false);
llama : support Llama 3 HF conversion (#6745) * Support Llama 3 conversion The tokenizer is BPE. * style * Accept suggestion Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> * llama : add llama_token_is_eog() ggml-ci * llama : auto-detect more EOT tokens when missing in KV data * convert : replacing EOS token is a hack * llama : fix codegemma EOT token + add TODOs * llama : fix model type string for 8B model --------- Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-21 11:50:41 +00:00
if (llama_token_is_eog(&ctx->model, id)) {
if (!allow_eog) {
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
candidates->data[i].logit = -INFINITY;
}
llama : more tokenizer fixes (#2810) * tests : write a Python tokenizer test (wip) * llama : prefix input text for tokenization with whitespace * llama : distinguish pieces from decoded text + fix detokenization * common : add comments * examples : no longer manually add leading space when tokenizing * tests : use Python to generate tokenizer tests for C++ * tests : add option to tokenize text files ggml-ci * tests : add test-tokenizer-1.py * llama.cpp : fix LF token * hellaswag : move the concat space for clarity * tests : add falcon tests (py + cpp, currently do not pass Unicode) ggml-ci * common : temporary separate llama_detokenize calls for SPM and BPE --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com>
2023-08-27 11:19:19 +00:00
} else if (piece.empty() || piece[0] == 0) {
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
candidates->data[i].logit = -INFINITY;
} else {
llama : grammar `reserve` space in `decode_utf8` (#4210) * reserve space for codepoints * improvement for the appended 0
2023-11-25 16:58:23 +00:00
candidates_decoded.push_back(decode_utf8(piece, grammar->partial_utf8));
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
candidates_grammar.push_back({ i, candidates_decoded.back().first.data(), candidates_decoded.back().second });
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
}
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
const auto rejects = llama_grammar_reject_candidates(grammar->rules, grammar->stacks, candidates_grammar);
for (const auto & reject : rejects) {
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
candidates->data[reject.index].logit = -INFINITY;
}
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
}
llama : add classifier-free guidance (#2135) * Initial implementation * Remove debug print * Restore signature of llama_init_from_gpt_params * Free guidance context * Make freeing of guidance_ctx conditional * Make Classifier-Free Guidance a sampling function * Correct typo. CFG already means context-free grammar. * Record sampling time in llama_sample_classifier_free_guidance * Shift all values by the max value before applying logsoftmax * Fix styling based on review
2023-07-11 16:18:43 +00:00
static void llama_log_softmax(float * array, size_t size) {
float max_l = *std::max_element(array, array + size);
float sum = 0.f;
for (size_t i = 0; i < size; ++i) {
float p = expf(array[i] - max_l);
sum += p;
array[i] = p;
}
for (size_t i = 0; i < size; ++i) {
array[i] = logf(array[i] / sum);
}
}
llama : apply classifier-free guidance to logits directly (#4951)
2024-01-15 13:06:52 +00:00
void llama_sample_apply_guidance(
struct llama_context * ctx,
float * logits,
float * logits_guidance,
float scale) {
GGML_ASSERT(ctx);
const auto t_start_sample_us = ggml_time_us();
const auto n_vocab = llama_n_vocab(llama_get_model(ctx));
llama_log_softmax(logits, n_vocab);
llama_log_softmax(logits_guidance, n_vocab);
for (int i = 0; i < n_vocab; ++i) {
auto & l = logits[i];
const auto & g = logits_guidance[i];
l = scale * (l - g) + g;
}
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
}
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
llama_token llama_sample_token_mirostat(struct llama_context * ctx, llama_token_data_array * candidates, float tau, float eta, int32_t m, float * mu) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
GGML_ASSERT(ctx);
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
auto N = float(llama_n_vocab(llama_get_model(ctx)));
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
int64_t t_start_sample_us;
t_start_sample_us = ggml_time_us();
llama_sample_softmax(nullptr, candidates);
// Estimate s_hat using the most probable m tokens
float s_hat = 0.0;
float sum_ti_bi = 0.0;
float sum_ti_sq = 0.0;
for (size_t i = 0; i < size_t(m - 1) && i < candidates->size - 1; ++i) {
float t_i = logf(float(i + 2) / float(i + 1));
float b_i = logf(candidates->data[i].p / candidates->data[i + 1].p);
sum_ti_bi += t_i * b_i;
sum_ti_sq += t_i * t_i;
}
s_hat = sum_ti_bi / sum_ti_sq;
// Compute k from the estimated s_hat and target surprise value
float epsilon_hat = s_hat - 1;
float k = powf((epsilon_hat * powf(2, *mu)) / (1 - powf(N, -epsilon_hat)), 1 / s_hat);
// Sample the next word X using top-k sampling
Remove default arguments from sampling functions (#1343)
2023-05-06 21:01:47 +00:00
llama_sample_top_k(nullptr, candidates, int(k), 1);
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
if (ctx) {
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
}
llama_token X = llama_sample_token(ctx, candidates);
t_start_sample_us = ggml_time_us();
// Compute error as the difference between observed surprise and target surprise value
size_t X_idx = std::distance(candidates->data, std::find_if(candidates->data, candidates->data + candidates->size, [&](const llama_token_data & candidate) {
return candidate.id == X;
}));
float observed_surprise = -log2f(candidates->data[X_idx].p);
float e = observed_surprise - tau;
// Update mu using the learning rate and error
*mu = *mu - eta * e;
if (ctx) {
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
}
return X;
}
llama_token llama_sample_token_mirostat_v2(struct llama_context * ctx, llama_token_data_array * candidates, float tau, float eta, float * mu) {
int64_t t_start_sample_us;
t_start_sample_us = ggml_time_us();
llama_sample_softmax(ctx, candidates);
// Truncate the words with surprise values greater than mu
candidates->size = std::distance(candidates->data, std::find_if(candidates->data, candidates->data + candidates->size, [&](const llama_token_data & candidate) {
return -log2f(candidate.p) > *mu;
}));
train : improved training-from-scratch example (#1652) * add python wrapper https://gist.github.com/abetlen/2b90e5f153f6efd00931d098de5c73ce * fix decoding error. adds errors=ignore parameter * add python bindings for functions to get and set the whole llama state (rng, logits, embedding and kv_cache) * update python bindings * add text generating baby-llama from scratch example * fix race condition bug in ggml_compute_forward_diag_mask_f32 * implement ggml_soft_max_back for more performant backward pass of soft_max avoids creating big intermediate matrices of size n_embd x n_embd for llama layers and n_vocab x n_vocab for cross entropy loss * improve softmax backward pass go from quadratic runtime to linear runtime by simplifying the formulas * fix race condition bug in non-inplace ggml_compute_forward_diag_mask_f32 memcpy needs to be synchronized across threads to avoid race conditions. => do it in INIT phase * fix bug in ggml_compute_forward_soft_max_back_f32 on DEBUG build * improve performance of mul_mat backward pass avoid transpose by using mul_mat with swapped arguments * avoid printing too much newlines in baby-llama-text * activate threading in baby-llama-text * add ggml_out_prod and use it for mul_mat backward pass for improved performance performance stats report improvement from 37 seconds to 16 seconds runtime during my training tests * better weight initialization improves training convergence at start * better weight initialization improves training convergence at start * improve ggml_out_prod performance - change iteration order (>15s -> 10s runtime) - parallelize over one more dimension: over dst matrix rows (10s -> <5s runtime) * add llama sampler, shuffle samples and constrain sampling to tokens occurring in train data * fix get_samples call, add model tensor names, increase model size, start training samples after newline * save train trained model to checkpoint and load model to be trained from checkpoint * use inplace functions where possible * initialize rng with srand * use different arguments for input and output checkpoint * ggml fixes to support backward pass on inplace operations * remove duplicate include * fix cross entropy loss - add target probabilities for each sample which is then used in cross entropy loss * print used memory before and after optimization * sample with non-greedy sampling parameters at the end of training * add cmake target for baby-llama-text * add ggml_add1_inplace to header * enable gradient propagation for inplace add1 and scale operations those functions backward passes don't need the original src0, so they also work when forward is inplace * implement AdamW in ggml_opt_adam by adding weight decay parameter (default 0.001f) also add a schedule parameter (default 1.0f) that can be used to scale alpha and decay according to learning schedule. setting the decay parameter to zero disables AdamW resulting in normal Adam optimizer. since the difference between Adam and AdamW is minimal it is not implemented as another optimizer, but integrated into the existing Adam optimizer. * use inplace operations in cross_entropy_loss * fix random weight initialization scale * add missing default parameters for adam optimizer * add ggml_opt_context, so that we can properly resume training otherwise the optimizer states, tracking statistics about the error function and its derivates, will reset to zero each time ggml_opt is called, hindering convergence on resumed training. now the optimizer context and all its memory is stored in a separate struct. * fix bug in llama_sample_token_mirostat_v2 when all candidates are filtered out through mu threshold, the following soft_max operation will fail. so keep at least one. * add forward function without using cache, for more performant training during training on whole samples no cache is required. removing the cache and simplifying the remaining code results in performance and memory usage improvement. * print suppressed newline tokens as string "\n" printing too much actual newlines is suppressed to avoid flooding the console. * store optimizer state in training checkpoint and add learning schedule persistent optimizer state allows to resume training without resetting the optimizer learning schedule consists of linear warmup ramp followed by cosine decay with restarts * remove unused functions * fix bug in get_samples which corrupted training targets * save checkpoint only when it was trained * simplify code * remove trailing whitespace * simplify backward pass for SQRT * replace inefficient repeat backward pass with dedicated repeat_back operation * add ggml_cross_entropy_loss with backward pass for faster training cross entropy loss can also be implemented using softmax and log, but as dedicated operation it is faster and especially avoids unnecessary memory overhead. * add tests for cross_entropy_loss backward pass finite differences regularly results in estimated gradient of zero, despite the backward pass giving non zero gradient. _probably_ the finite differences fails due to numerical issues * use ggml_cross_entropy_loss in text training example * remove trailing whitespace * slightly improve how cross entropy loss is compute btw: directly implemented cross entropy loss seems to have way lower magnitudes than when implemented with softmax and log. probably the input to log gets closer to zero due to float numerics. maybe the multiplication by (1.0-eps)/sum is more accurate.. * add llama_get_vocab to get the vocabulary as output parameters * set default model.type for unknown models with few layers * add export of training checkpoint to llama compatible model file * get vocabulary for exporting training checkpoint to llama compatible model file * implement backward pass of flash attention * bugfixes for backward pass of flash attention * test flash attention backward pass need to set loose error bounds to pass. the finitie differences are close to numeric limits and often return quite different values than the backward pass. reducing eps further lets the gradients vanish completely. likewise setting eps to big results in wronger values. the softmax in the middle of the function is probably the most responsible for the numeric issues using finite differences. * add option to train with flash attention and move options to the top of the main function training from scratch also works with flash attention training convergence and generation results after fix number of iterations are worse than when not using flash attention. maybe there still lingers a bug in the flash attention backward pass? but training works, just with slower convergence. flash attention is still worth to use, because it requires way less memory and is faster with high n_ctx * add train_params and command line option parser * remove unnecessary comments * add train params to specify memory size * remove python bindings * rename baby-llama-text to train-text-from-scratch * replace auto parameters in lambda function * add #include <climits> * add explicit cast to fix compile error "error: non-constant-expression cannot be narrowed from type 'int64_t' (aka 'long long') to 'uint32_t' (aka 'unsigned int') in initializer list [-Wc++11-narrowing]" * remove trailing whitespace * add ggml_opt_resume_g which accepts forward and backward cgraphs * fix formulas in comments * bug fix for ggml_compute_forward_get_rows_back_f32 the result should be set to zero, not to whatever data is in opt0 * improve training memory usage with scratch buffers instead of relying on the automatic backward pass, we manually create the graph for the backward pass. it turns out that all backward pass operations need only temporary memory which can be reused after each layer. will compute backward pass for ALL model parameters * add option to use scratch buffers in training or not make it configurable because currently training with scratch buffers implies flash attention and optimization over all parameters. * ci : disable temporary * store view offset and permute axes in opt[0] instead of storing it in padding use memcpy to store offset, because offset is of type size_t. when storing it as int32_t offset would have to be smaller than 2^31 which is not necessarily true. * minor : fix compile warnings + minor style changes * fix bug in threaded indices calculation of ggml_compute_forward_flash_attn_back_f32 * store view offset like in master branch * bug fix in forward_batch_wo_cache_flash_attn_train * scratch buffer bug fixes in forward_batch_wo_cache_flash_attn_train data of permute and reshape is the same as their input. if we want to preserve the output of permute/reshape, we also need to preserve their inputs. replace reshape(src0, src1) with reshape_nd calls so that we don't need src1. replace (temporary) t03 with ggml_repeat(ctx0, layer.attention_norm, t02). in the future we could also use the new broadcasting ggml_mul to avoid these repeat calls. for this we need backward pass of broadcasting ggml_mul. * remove unnecessary scratch buffer 0 buf 0 is persistent memory, so we can just disable scratch for this by using buf -1 * avoid creating unnecessary grad tensors previously we need to create grads for model parameters, so that expand(..) correctly populates cgraph->leafs & cgraph->grads this wasted memory, because unnecessary grad for each op were automatically created: the automatically generated grad was unnecessary because we later manually set the grad (e.g. t35->grad = expand(gb, ...) ). this discarded the automatically generated grad resulting in wasted memory. improved this by changing expand(..) to not use ggml_build_forward_expand. expand set cgraph->nodes but not the leafs. cgraph->leafs & cgraph->grads are set in another pass after the last expand call. * print used training seed * zero initialize gfbuf and gbbuf * ci : re-enable workflows + add README for training --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-13 19:04:40 +00:00
if (candidates->size == 0) {
candidates->size = 1;
}
llama: Don't double count the sampling time (#2107)
2023-07-05 10:31:23 +00:00
if (ctx) {
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
}
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
// Normalize the probabilities of the remaining words
llama_sample_softmax(ctx, candidates);
// Sample the next word X from the remaining words
llama_token X = llama_sample_token(ctx, candidates);
t_start_sample_us = ggml_time_us();
// Compute error as the difference between observed surprise and target surprise value
size_t X_idx = std::distance(candidates->data, std::find_if(candidates->data, candidates->data + candidates->size, [&](const llama_token_data & candidate) {
return candidate.id == X;
}));
float observed_surprise = -log2f(candidates->data[X_idx].p);
float e = observed_surprise - tau;
// Update mu using the learning rate and error
*mu = *mu - eta * e;
if (ctx) {
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
}
return X;
}
llama_token llama_sample_token_greedy(struct llama_context * ctx, llama_token_data_array * candidates) {
const int64_t t_start_sample_us = ggml_time_us();
// Find max element
llama : fix various warnings
2023-05-13 08:23:15 +00:00
auto * max_iter = std::max_element(candidates->data, candidates->data + candidates->size, [](const llama_token_data & a, const llama_token_data & b) {
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
return a.logit < b.logit;
});
llama_token result = max_iter->id;
if (ctx) {
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
ctx->n_sample++;
}
return result;
}
Server: fix seed for multiple slots (#6835) * Server: add tests for consistent results * sampling: separate rng per sampling context
2024-04-24 09:08:36 +00:00
llama_token llama_sample_token_with_rng(struct llama_context * ctx, llama_token_data_array * candidates, std::mt19937 & rng) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
GGML_ASSERT(ctx);
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
const int64_t t_start_sample_us = ggml_time_us();
llama_sample_softmax(nullptr, candidates);
std::vector<float> probs;
probs.reserve(candidates->size);
for (size_t i = 0; i < candidates->size; ++i) {
probs.push_back(candidates->data[i].p);
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
std::discrete_distribution<> dist(probs.begin(), probs.end());
int idx = dist(rng);
llama : new sampling algorithms (#1126) * Sample interface, new samplers. New samplers: - locally typical sampling - tail free sampling - frequency and presence penalty - mirostat Ignore EOS fix: -inf should be used. * mirostat * Added --logit-bias and --no-penalize-nl, removed std::span * Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) Use C++11, clarify llama API documentation, rename Mirostat parameters to --mirostat_lr and --mirostat_ent, add temperature sampling for Mirostat, simplify Mirostat sampling API parameters (removed N and *k) * Save and load example adjust * Tests * Windows build fix * Windows test fix
2023-04-29 05:34:41 +00:00
llama_token result = candidates->data[idx].id;
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
ctx->n_sample++;
return result;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
Server: fix seed for multiple slots (#6835) * Server: add tests for consistent results * sampling: separate rng per sampling context
2024-04-24 09:08:36 +00:00
llama_token llama_sample_token(struct llama_context * ctx, llama_token_data_array * candidates) {
return llama_sample_token_with_rng(ctx, candidates, ctx->rng);
}
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
void llama_grammar_accept_token(struct llama_context * ctx, struct llama_grammar * grammar, llama_token token) {
const int64_t t_start_sample_us = ggml_time_us();
llama : support Llama 3 HF conversion (#6745) * Support Llama 3 conversion The tokenizer is BPE. * style * Accept suggestion Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> * llama : add llama_token_is_eog() ggml-ci * llama : auto-detect more EOT tokens when missing in KV data * convert : replacing EOS token is a hack * llama : fix codegemma EOT token + add TODOs * llama : fix model type string for 8B model --------- Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-21 11:50:41 +00:00
if (llama_token_is_eog(&ctx->model, token)) {
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
for (const auto & stack : grammar->stacks) {
if (stack.empty()) {
return;
}
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
GGML_ASSERT(false);
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
}
llama : add option to render special/control tokens (#6807) * make : fix common dep on llama.h * llama : add option to render special tokens * readme : add API change notice ggml-ci * swift : fix build
2024-04-21 15:36:45 +00:00
const std::string piece = llama_token_to_piece(ctx, token, false);
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
// Note terminating 0 in decoded string
llama : grammar `reserve` space in `decode_utf8` (#4210) * reserve space for codepoints * improvement for the appended 0
2023-11-25 16:58:23 +00:00
const auto decoded = decode_utf8(piece, grammar->partial_utf8);
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
const auto & code_points = decoded.first;
grammars: 1.5x faster inference w/ complex grammars (vector reserves / reuses) (#6609) * grammars: reserve rejects & next candidates * grammars: reuse new_stacks * grammars: fix missing sig change in llama.h * grammars: fix test (api changed) * grammars: update gbnf-validator.cpp * grammars: simpler syntax (no swap)
2024-04-11 18:47:34 +00:00
std::vector<std::vector<const llama_grammar_element *>> tmp_new_stacks;
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
for (auto it = code_points.begin(), end = code_points.end() - 1; it != end; ++it) {
grammars: 1.5x faster inference w/ complex grammars (vector reserves / reuses) (#6609) * grammars: reserve rejects & next candidates * grammars: reuse new_stacks * grammars: fix missing sig change in llama.h * grammars: fix test (api changed) * grammars: update gbnf-validator.cpp * grammars: simpler syntax (no swap)
2024-04-11 18:47:34 +00:00
llama_grammar_accept(grammar->rules, grammar->stacks, *it, tmp_new_stacks);
grammar->stacks = tmp_new_stacks;
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
}
Fix unicode in grammars (fixes #2501) (#2553) * Fix unicode in grammars (fixes #2501) * add more comments * fix test-llama-grammar
2023-08-17 23:54:44 +00:00
grammar->partial_utf8 = decoded.second;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
GGML_ASSERT(!grammar->stacks.empty());
llama : add grammar-based sampling (#1773) * llama, main : constrain sampling to grammar * allow loading grammar from file * fix whitespace errors * handle & print parser errors * add comments to grammar syntax and allow newlines where unambiguous * add missing include * support alternates in root rule * fix bugs with empty token and EOS * adjust JSON grammar * remove swp file * rewrite ternary expressions Co-authored-by: Henri Vasserman <henv@hot.ee> * use struct for grammar elements and add Unicode support * add unicode escapes * add inverse char ranges * only sample full tokens (no peeking or truncation) * llama : minor style changes blindly applied in online editor - hopefully I didn't break something * update help text * add warning message if EOS is disabled --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-24 03:58:10 +00:00
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
}
llama : add llama_beam_search() (#2267) * Add llama_beam_search(). * Add '// Beam search' heading to llama.{h,cpp} after llama_grammar_accept_token(). * Add space around * pointers and & references. * Add spaces around comparison and assignment operators. * Prefer west const. * Use llama_ prefix for structs in global namespace. * Delete obsolete comment from an earlier revision. * Change eos to eob in llama_beam and llama_beam_view structs.
2023-08-25 15:18:48 +00:00
//
// Beam search
//
struct llama_beam {
std::vector<llama_token> tokens;
float p; // Cumulative beam probability (renormalized relative to all beams)
bool eob; // Initialize end-of-beam to false. Callback sets this to true.
// Sort beams by probability. In case of ties, prefer beams at eob.
bool operator<(const llama_beam & rhs) const {
return std::make_pair(p, eob) < std::make_pair(rhs.p, rhs.eob);
}
// Shift off first n tokens and discard them.
void shift_tokens(const size_t n) {
if (n) {
std::copy(tokens.begin() + n, tokens.end(), tokens.begin());
tokens.resize(tokens.size() - n);
}
}
llama_beam_view view() const { return {tokens.data(), tokens.size(), p, eob}; }
};
// A struct for calculating logit-related info.
struct llama_logit_info {
const float * const logits;
const int n_vocab;
const float max_l;
const float normalizer;
struct sum_exp {
float max_l;
float operator()(float sum, float l) const { return sum + std::exp(l - max_l); }
};
llama_logit_info(llama_context * ctx)
: logits(llama_get_logits(ctx))
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
, n_vocab(llama_n_vocab(llama_get_model(ctx)))
llama : add llama_beam_search() (#2267) * Add llama_beam_search(). * Add '// Beam search' heading to llama.{h,cpp} after llama_grammar_accept_token(). * Add space around * pointers and & references. * Add spaces around comparison and assignment operators. * Prefer west const. * Use llama_ prefix for structs in global namespace. * Delete obsolete comment from an earlier revision. * Change eos to eob in llama_beam and llama_beam_view structs.
2023-08-25 15:18:48 +00:00
, max_l(*std::max_element(logits, logits + n_vocab))
, normalizer(1.0f / std::accumulate(logits, logits + n_vocab, 0.0f, sum_exp{max_l}))
{ }
llama_token_data get_token_data(const llama_token token_id) const {
constexpr auto p = std::numeric_limits<float>::quiet_NaN(); // never used
return {token_id, logits[token_id], p};
}
// Return top k token_data by logit.
std::vector<llama_token_data> top_k(size_t k) {
std::vector<llama_token_data> min_heap; // min-heap by logit
const llama_token k_min = std::min(static_cast<llama_token>(k), n_vocab);
min_heap.reserve(k_min);
for (llama_token token_id = 0 ; token_id < k_min ; ++token_id) {
min_heap.push_back(get_token_data(token_id));
}
auto comp = [](const llama_token_data & a, const llama_token_data & b) { return a.logit > b.logit; };
std::make_heap(min_heap.begin(), min_heap.end(), comp);
for (llama_token token_id = k_min ; token_id < n_vocab ; ++token_id) {
if (min_heap.front().logit < logits[token_id]) {
std::pop_heap(min_heap.begin(), min_heap.end(), comp);
min_heap.back().id = token_id;
min_heap.back().logit = logits[token_id];
std::push_heap(min_heap.begin(), min_heap.end(), comp);
}
}
return min_heap;
}
minor : add const qualifiers (#2853) * made the methods const # Conflicts: # examples/convert-llama2c-to-ggml/convert-llama2c-to-ggml.cpp * made method const * Update convert-llama2c-to-ggml.cpp removed write_raw and write_u32 * llama2c : remove misleading const --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-01 13:47:27 +00:00
float probability_from_logit(float logit) const {
llama : add llama_beam_search() (#2267) * Add llama_beam_search(). * Add '// Beam search' heading to llama.{h,cpp} after llama_grammar_accept_token(). * Add space around * pointers and & references. * Add spaces around comparison and assignment operators. * Prefer west const. * Use llama_ prefix for structs in global namespace. * Delete obsolete comment from an earlier revision. * Change eos to eob in llama_beam and llama_beam_view structs.
2023-08-25 15:18:48 +00:00
return normalizer * std::exp(logit - max_l);
}
};
struct llama_beam_search_data {
llama_context * ctx;
size_t n_beams;
int n_past;
int n_predict;
std::vector<llama_beam> beams;
std::vector<llama_beam> next_beams;
// Re-calculated on each loop iteration
size_t common_prefix_length;
// Used to communicate to/from callback on beams state.
std::vector<llama_beam_view> beam_views;
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
llama_beam_search_data(llama_context * ctx, size_t n_beams, int n_past, int n_predict)
llama : add llama_beam_search() (#2267) * Add llama_beam_search(). * Add '// Beam search' heading to llama.{h,cpp} after llama_grammar_accept_token(). * Add space around * pointers and & references. * Add spaces around comparison and assignment operators. * Prefer west const. * Use llama_ prefix for structs in global namespace. * Delete obsolete comment from an earlier revision. * Change eos to eob in llama_beam and llama_beam_view structs.
2023-08-25 15:18:48 +00:00
: ctx(ctx)
, n_beams(n_beams)
, n_past(n_past)
, n_predict(n_predict)
, beam_views(n_beams) {
beams.reserve(n_beams);
next_beams.reserve(n_beams);
}
// Collapse beams to a single beam given by index.
void collapse_beams(const size_t beam_idx) {
if (0u < beam_idx) {
std::swap(beams[0], beams[beam_idx]);
}
beams.resize(1);
}
// Min-heaps are used to efficiently collect the top-k elements (k=n_beams).
english : use `typos` to fix comments and logs (#4354)
2023-12-12 09:53:36 +00:00
// The repetitive patterns below reflect the 2 stages of heaps:
llama : add llama_beam_search() (#2267) * Add llama_beam_search(). * Add '// Beam search' heading to llama.{h,cpp} after llama_grammar_accept_token(). * Add space around * pointers and & references. * Add spaces around comparison and assignment operators. * Prefer west const. * Use llama_ prefix for structs in global namespace. * Delete obsolete comment from an earlier revision. * Change eos to eob in llama_beam and llama_beam_view structs.
2023-08-25 15:18:48 +00:00
// * Gather elements until the vector is full, then call std::make_heap() on it.
// * If the heap is full and a new element is found that should be included, pop the
// least element to the back(), replace it with the new, then push it into the heap.
void fill_next_beams_by_top_probabilities(llama_beam & beam) {
// Min-heaps use a greater-than comparator.
const auto comp = [](const llama_beam & a, const llama_beam & b) { return a.p > b.p; };
if (beam.eob) {
// beam is at end-of-sentence, so just copy it to next_beams if its probability is high enough.
if (next_beams.size() < n_beams) {
next_beams.push_back(std::move(beam));
if (next_beams.size() == n_beams) {
std::make_heap(next_beams.begin(), next_beams.end(), comp);
}
} else if (next_beams.front().p < beam.p) {
std::pop_heap(next_beams.begin(), next_beams.end(), comp);
next_beams.back() = std::move(beam);
std::push_heap(next_beams.begin(), next_beams.end(), comp);
}
} else {
// beam is not at end-of-sentence, so branch with next top_k tokens.
if (!beam.tokens.empty()) {
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
llama_decode(ctx, llama_batch_get_one(beam.tokens.data(), beam.tokens.size(), n_past, 0));
llama : add llama_beam_search() (#2267) * Add llama_beam_search(). * Add '// Beam search' heading to llama.{h,cpp} after llama_grammar_accept_token(). * Add space around * pointers and & references. * Add spaces around comparison and assignment operators. * Prefer west const. * Use llama_ prefix for structs in global namespace. * Delete obsolete comment from an earlier revision. * Change eos to eob in llama_beam and llama_beam_view structs.
2023-08-25 15:18:48 +00:00
}
llama_logit_info logit_info(ctx);
std::vector<llama_token_data> next_tokens = logit_info.top_k(n_beams);
llama : add missing kv clear in llama_beam_search (#6664)
2024-04-14 19:24:15 +00:00
// Clear the kv slot so that other beams may try different tokens at this position. The llama_decode()
// call in loop() will conclusively fill in the kv slot once the beams converge at this position.
llama_kv_cache_seq_rm(ctx, 0, n_past, -1);
llama : add llama_beam_search() (#2267) * Add llama_beam_search(). * Add '// Beam search' heading to llama.{h,cpp} after llama_grammar_accept_token(). * Add space around * pointers and & references. * Add spaces around comparison and assignment operators. * Prefer west const. * Use llama_ prefix for structs in global namespace. * Delete obsolete comment from an earlier revision. * Change eos to eob in llama_beam and llama_beam_view structs.
2023-08-25 15:18:48 +00:00
size_t i=0;
if (next_beams.size() < n_beams) {
for (; next_beams.size() < n_beams ; ++i) {
llama_beam next_beam = beam;
next_beam.tokens.push_back(next_tokens[i].id);
next_beam.p *= logit_info.probability_from_logit(next_tokens[i].logit);
next_beams.push_back(std::move(next_beam));
}
std::make_heap(next_beams.begin(), next_beams.end(), comp);
} else {
for (; next_beams.front().p == 0.0f ; ++i) {
std::pop_heap(next_beams.begin(), next_beams.end(), comp);
next_beams.back() = beam;
next_beams.back().tokens.push_back(next_tokens[i].id);
next_beams.back().p *= logit_info.probability_from_logit(next_tokens[i].logit);
std::push_heap(next_beams.begin(), next_beams.end(), comp);
}
}
for (; i < n_beams ; ++i) {
const float next_p = beam.p * logit_info.probability_from_logit(next_tokens[i].logit);
if (next_beams.front().p < next_p) {
std::pop_heap(next_beams.begin(), next_beams.end(), comp);
next_beams.back() = beam;
next_beams.back().tokens.push_back(next_tokens[i].id);
next_beams.back().p = next_p;
std::push_heap(next_beams.begin(), next_beams.end(), comp);
}
}
}
}
// Find common_prefix_length based on beams.
// Requires beams is not empty.
size_t find_common_prefix_length() {
size_t common_prefix_length = beams[0].tokens.size();
for (size_t i = 1 ; i < beams.size() ; ++i) {
common_prefix_length = std::min(common_prefix_length, beams[i].tokens.size());
for (size_t j = 0 ; j < common_prefix_length ; ++j) {
if (beams[0].tokens[j] != beams[i].tokens[j]) {
common_prefix_length = j;
break;
}
}
}
return common_prefix_length;
}
// Construct beams_state to send back to caller via the callback function.
// Side effect: set common_prefix_length = find_common_prefix_length();
llama_beams_state get_beams_state(const bool last_call) {
for (size_t i = 0 ; i < beams.size() ; ++i) {
beam_views[i] = beams[i].view();
}
common_prefix_length = find_common_prefix_length();
return {beam_views.data(), beams.size(), common_prefix_length, last_call};
}
// Loop:
// * while i < n_predict, AND
// * any of the beams have not yet reached end-of-beam (eob), AND
// * the highest probability beam(s) (plural in case of ties) are not at end-of-sentence
// (since all other beam probabilities can only decrease)
void loop(const llama_beam_search_callback_fn_t callback, void * const callback_data) {
beams.push_back({{}, 1.0f, false}); // Start with one empty beam w/ probability = 1.0 and !eob.
const auto not_eob = [](const llama_beam & beam) { return !beam.eob; };
for (int i = 0 ; i < n_predict && std::any_of(beams.begin(),beams.end(),not_eob) &&
!beams[top_beam_index()].eob ; ++i) {
callback(callback_data, get_beams_state(false)); // Sets common_prefix_length
update_beams_from_beam_views(); // Update values (p,eob) that callback may have changed.
if (common_prefix_length) {
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
llama_decode(ctx, llama_batch_get_one(beams[0].tokens.data(), common_prefix_length, n_past, 0));
llama : add llama_beam_search() (#2267) * Add llama_beam_search(). * Add '// Beam search' heading to llama.{h,cpp} after llama_grammar_accept_token(). * Add space around * pointers and & references. * Add spaces around comparison and assignment operators. * Prefer west const. * Use llama_ prefix for structs in global namespace. * Delete obsolete comment from an earlier revision. * Change eos to eob in llama_beam and llama_beam_view structs.
2023-08-25 15:18:48 +00:00
n_past += common_prefix_length;
}
// Zero-out next_beam probabilities to place them last in following min-heap.
std::for_each(next_beams.begin(), next_beams.end(), [](llama_beam & beam) { beam.p = 0.0f; });
for (llama_beam & beam : beams) {
beam.shift_tokens(common_prefix_length);
fill_next_beams_by_top_probabilities(beam);
}
// next_beams become the beams of next/final iteration. Swap them to re-use memory.
beams.swap(next_beams);
renormalize_beam_probabilities(beams);
}
collapse_beams(top_beam_index());
callback(callback_data, get_beams_state(true));
}
// As beams grow, the cumulative probabilities decrease.
// Renormalize them to avoid floating point underflow.
static void renormalize_beam_probabilities(std::vector<llama_beam> & beams) {
const auto sum_p = [](float sum, llama_beam & beam) { return sum + beam.p; };
const float inv_sum = 1.0f / std::accumulate(beams.begin(), beams.end(), 0.0f, sum_p);
std::for_each(beams.begin(), beams.end(), [=](llama_beam & beam) { beam.p *= inv_sum; });
}
// Assumes beams is non-empty. Uses llama_beam::operator<() for ordering.
size_t top_beam_index() {
return std::max_element(beams.begin(), beams.end()) - beams.begin();
}
// Copy (p,eob) for each beam which may have been changed by the callback.
void update_beams_from_beam_views() {
for (size_t i = 0 ; i < beams.size() ; ++i) {
beams[i].p = beam_views[i].p;
beams[i].eob = beam_views[i].eob;
}
}
};
void llama_beam_search(llama_context * ctx,
llama_beam_search_callback_fn_t callback, void * callback_data,
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
size_t n_beams, int n_past, int n_predict) {
llama : add llama_beam_search() (#2267) * Add llama_beam_search(). * Add '// Beam search' heading to llama.{h,cpp} after llama_grammar_accept_token(). * Add space around * pointers and & references. * Add spaces around comparison and assignment operators. * Prefer west const. * Use llama_ prefix for structs in global namespace. * Delete obsolete comment from an earlier revision. * Change eos to eob in llama_beam and llama_beam_view structs.
2023-08-25 15:18:48 +00:00
assert(ctx);
const int64_t t_start_sample_us = ggml_time_us();
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
llama_beam_search_data beam_search_data(ctx, n_beams, n_past, n_predict);
llama : add llama_beam_search() (#2267) * Add llama_beam_search(). * Add '// Beam search' heading to llama.{h,cpp} after llama_grammar_accept_token(). * Add space around * pointers and & references. * Add spaces around comparison and assignment operators. * Prefer west const. * Use llama_ prefix for structs in global namespace. * Delete obsolete comment from an earlier revision. * Change eos to eob in llama_beam and llama_beam_view structs.
2023-08-25 15:18:48 +00:00
beam_search_data.loop(callback, callback_data);
ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
ctx->n_sample++;
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
//
// quantization
//
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
struct quantize_state_internal {
const llama_model & model;
const llama_model_quantize_params * params;
ggml : quantization refactoring (#3833) * ggml : factor all quantization code in ggml-quants ggml-ci * ggml-quants : fix Zig and Swift builds + quantize tool ggml-ci * quantize : --pure option for disabling k-quant mixtures --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com>
2023-10-29 16:32:28 +00:00
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
int n_attention_wv = 0;
llama : add Q3_K_XS (#5060) * Add Q3_K_XS - intermediate size between Q2_K and Q3_K_S * Q3_K_XS: quanize first 1/8 of ffn_down layers with Q4_K Together with an importance matrix, this brings perplexity for LLaMA-v2-70B below the perplexity of the former Q2_K with a 800 MB smaller quantized model size. --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-22 10:43:33 +00:00
int n_ffn_down = 0;
int n_ffn_gate = 0;
int n_ffn_up = 0;
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
int i_attention_wv = 0;
llama : add Q3_K_XS (#5060) * Add Q3_K_XS - intermediate size between Q2_K and Q3_K_S * Q3_K_XS: quanize first 1/8 of ffn_down layers with Q4_K Together with an importance matrix, this brings perplexity for LLaMA-v2-70B below the perplexity of the former Q2_K with a 800 MB smaller quantized model size. --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-22 10:43:33 +00:00
int i_ffn_down = 0;
int i_ffn_gate = 0;
int i_ffn_up = 0;
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
int n_k_quantized = 0;
int n_fallback = 0;
ggml : quantization refactoring (#3833) * ggml : factor all quantization code in ggml-quants ggml-ci * ggml-quants : fix Zig and Swift builds + quantize tool ggml-ci * quantize : --pure option for disabling k-quant mixtures --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com>
2023-10-29 16:32:28 +00:00
ggml : importance matrix support for legacy quants (#4969) * imatrix: adding support for legacy quants * imatrix: guard Q4_0/Q5_0 against ffn_down craziness --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-16 17:51:26 +00:00
bool has_imatrix = false;
llama : fix quantization of shared token_embd (#5944)
2024-03-08 15:53:37 +00:00
// used to figure out if a model shares tok_embd with the output weight
bool has_output = false;
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
quantize_state_internal(const llama_model & model, const llama_model_quantize_params * params)
: model(model)
, params(params)
{}
};
llama : refactor internal quantization functions (#5830)
2024-03-02 14:19:09 +00:00
static void llama_tensor_dequantize_internal(
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
struct ggml_tensor * tensor, std::vector<no_init<float>> & output, std::vector<std::thread> & workers,
const size_t nelements, const int nthread
) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
if (output.size() < nelements) {
output.resize(nelements);
llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691) * Add support for quantizing already quantized models * Threaded dequantizing and f16 to f32 conversion * Clean up thread blocks with spares calculation a bit * Use std::runtime_error exceptions.
2023-06-10 07:59:17 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
float * f32_output = (float *) output.data();
llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691) * Add support for quantizing already quantized models * Threaded dequantizing and f16 to f32 conversion * Clean up thread blocks with spares calculation a bit * Use std::runtime_error exceptions.
2023-06-10 07:59:17 +00:00
ggml : generalize `quantize_fns` for simpler FP16 handling (#1237) * Generalize quantize_fns for simpler FP16 handling * Remove call to ggml_cuda_mul_mat_get_wsize * ci : disable FMA for mac os actions --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-05 16:13:06 +00:00
ggml_type_traits_t qtype;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
if (ggml_is_quantized(tensor->type)) {
qtype = ggml_internal_get_type_traits(tensor->type);
ggml : generalize `quantize_fns` for simpler FP16 handling (#1237) * Generalize quantize_fns for simpler FP16 handling * Remove call to ggml_cuda_mul_mat_get_wsize * ci : disable FMA for mac os actions --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-05 16:13:06 +00:00
if (qtype.to_float == NULL) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
throw std::runtime_error(format("type %s unsupported for integer quantization: no dequantization available", ggml_type_name(tensor->type)));
llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691) * Add support for quantizing already quantized models * Threaded dequantizing and f16 to f32 conversion * Clean up thread blocks with spares calculation a bit * Use std::runtime_error exceptions.
2023-06-10 07:59:17 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
} else if (tensor->type != GGML_TYPE_F16) {
throw std::runtime_error(format("cannot dequantize/convert tensor type %s", ggml_type_name(tensor->type)));
llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691) * Add support for quantizing already quantized models * Threaded dequantizing and f16 to f32 conversion * Clean up thread blocks with spares calculation a bit * Use std::runtime_error exceptions.
2023-06-10 07:59:17 +00:00
}
if (nthread < 2) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
if (tensor->type == GGML_TYPE_F16) {
ggml_fp16_to_fp32_row((ggml_fp16_t *)tensor->data, f32_output, nelements);
} else if (ggml_is_quantized(tensor->type)) {
qtype.to_float(tensor->data, f32_output, nelements);
llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691) * Add support for quantizing already quantized models * Threaded dequantizing and f16 to f32 conversion * Clean up thread blocks with spares calculation a bit * Use std::runtime_error exceptions.
2023-06-10 07:59:17 +00:00
} else {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
GGML_ASSERT(false); // unreachable
llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691) * Add support for quantizing already quantized models * Threaded dequantizing and f16 to f32 conversion * Clean up thread blocks with spares calculation a bit * Use std::runtime_error exceptions.
2023-06-10 07:59:17 +00:00
}
return;
}
llama : fix integer overflow during quantization (#4284) happens with multi-threaded quantization of Qwen-72B ggml-ci
2023-12-01 16:42:11 +00:00
size_t block_size = tensor->type == GGML_TYPE_F16 ? 1 : (size_t)ggml_blck_size(tensor->type);
size_t block_size_bytes = ggml_type_size(tensor->type);
llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691) * Add support for quantizing already quantized models * Threaded dequantizing and f16 to f32 conversion * Clean up thread blocks with spares calculation a bit * Use std::runtime_error exceptions.
2023-06-10 07:59:17 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
GGML_ASSERT(nelements % block_size == 0);
llama : fix integer overflow during quantization (#4284) happens with multi-threaded quantization of Qwen-72B ggml-ci
2023-12-01 16:42:11 +00:00
size_t nblocks = nelements / block_size;
size_t blocks_per_thread = nblocks / nthread;
size_t spare_blocks = nblocks - (blocks_per_thread * nthread); // if blocks aren't divisible by thread count
size_t in_buff_offs = 0;
size_t out_buff_offs = 0;
for (int tnum = 0; tnum < nthread; tnum++) {
size_t thr_blocks = blocks_per_thread + (tnum == nthread - 1 ? spare_blocks : 0); // num blocks for this thread
size_t thr_elems = thr_blocks * block_size; // number of elements for this thread
size_t thr_block_bytes = thr_blocks * block_size_bytes; // number of input bytes for this thread
llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691) * Add support for quantizing already quantized models * Threaded dequantizing and f16 to f32 conversion * Clean up thread blocks with spares calculation a bit * Use std::runtime_error exceptions.
2023-06-10 07:59:17 +00:00
auto compute = [qtype] (ggml_type typ, uint8_t * inbuf, float * outbuf, int nels) {
if (typ == GGML_TYPE_F16) {
ggml_fp16_to_fp32_row((ggml_fp16_t *)inbuf, outbuf, nels);
} else {
ggml : generalize `quantize_fns` for simpler FP16 handling (#1237) * Generalize quantize_fns for simpler FP16 handling * Remove call to ggml_cuda_mul_mat_get_wsize * ci : disable FMA for mac os actions --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-05 16:13:06 +00:00
qtype.to_float(inbuf, outbuf, nels);
llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691) * Add support for quantizing already quantized models * Threaded dequantizing and f16 to f32 conversion * Clean up thread blocks with spares calculation a bit * Use std::runtime_error exceptions.
2023-06-10 07:59:17 +00:00
}
};
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
workers.emplace_back(compute, tensor->type, (uint8_t *) tensor->data + in_buff_offs, f32_output + out_buff_offs, thr_elems);
llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691) * Add support for quantizing already quantized models * Threaded dequantizing and f16 to f32 conversion * Clean up thread blocks with spares calculation a bit * Use std::runtime_error exceptions.
2023-06-10 07:59:17 +00:00
in_buff_offs += thr_block_bytes;
out_buff_offs += thr_elems;
}
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
for (auto & w : workers) { w.join(); }
workers.clear();
}
llama : fix F16/F32 downcast + improve names (#5980)
2024-03-11 07:56:47 +00:00
static ggml_type llama_tensor_get_type(quantize_state_internal & qs, ggml_type new_type, const ggml_tensor * tensor, llama_ftype ftype) {
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
const std::string name = ggml_get_name(tensor);
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
// TODO: avoid hardcoded tensor names - use the TN_* constants
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
const llm_arch arch = qs.model.arch;
const auto tn = LLM_TN(arch);
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
auto use_more_bits = [](int i_layer, int num_layers) -> bool {
return i_layer < num_layers/8 || i_layer >= 7*num_layers/8 || (i_layer - num_layers/8)%3 == 2;
};
Fix Q3_K_XS for MoE models (#5113) Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-25 15:58:53 +00:00
const int n_expert = std::max(1, (int)qs.model.hparams.n_expert);
auto layer_info = [n_expert] (int i_layer, int n_layer, const char * name) {
if (n_expert > 1) {
// Believe it or not, "experts" in the FFN of Mixtral-8x7B are not consecutive, but iccasionally randomly
// sprinkled in the model. Hence, simply dividing i_ffn_down by n_expert does not work
// for getting the current layer as I initially thought, and we need to resort to parsing the
// tensor name.
if (sscanf(name, "blk.%d.", &i_layer) != 1) {
throw std::runtime_error(format("Failed to determine layer for tensor %s", name));
}
if (i_layer < 0 || i_layer >= n_layer) {
throw std::runtime_error(format("Bad layer %d for tensor %s. Must be in [0, %d)", i_layer, name, n_layer));
}
}
return std::make_pair(i_layer, n_layer);
};
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
gemma : use more bits for the token_embd.weight tensor (#5650) * gemma : use Q8_0 for the token_embd.weight tensor * llama : quantize token_embd.weight using output type
2024-02-22 21:23:46 +00:00
// for arches that share the same tensor between the token embeddings and the output, we quantize the token embeddings
// with the quantization of the output tensor
llama : fix quantization of shared token_embd (#5944)
2024-03-08 15:53:37 +00:00
if (name == tn(LLM_TENSOR_OUTPUT, "weight") || (!qs.has_output && name == tn(LLM_TENSOR_TOKEN_EMBD, "weight"))) {
quantize: options for output and token embedding tensors qtype (#6239) * quantize: be able to specify the output tensor type * quantize: be able to specify the token embedding tensor type --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-22 18:47:14 +00:00
if (qs.params->output_tensor_type < GGML_TYPE_COUNT) {
new_type = qs.params->output_tensor_type;
} else {
int nx = tensor->ne[0];
if (arch == LLM_ARCH_FALCON || nx % QK_K != 0) {
new_type = GGML_TYPE_Q8_0;
}
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ2_XXS || ftype == LLAMA_FTYPE_MOSTLY_IQ2_XS || ftype == LLAMA_FTYPE_MOSTLY_IQ3_XXS ||
IQ1_M: 1.75 bpw quantization (#6302) * iq1_m: basics * iq1_m: basics-2 * iq1_m: CUDA dequantize works Very 1st shot I get PPL = 9.76 for LLaMA-v2-7B. * iq1_m: separate shifts for each group of 8 in a block We get PPL(LLaMA-v2-7B ) = 9.2810 PPL(LLaMA-v2-13B) = 6.8105 Not bad, but slightly higher than sqrt(PPL(IQ1_S) * PPL(IQ2_XXS)) which is the expected outcome given that IQ1_M is halfway between IQ1_S and IQ2_XXS in terms of bpw. From this, we would expect PPL = 9.14 for LLaMA-v2-7B PPL = 6.63 for LLaMA-v2-13B * iq1_m: go to 3-bit scales There is slight increase in PPL, but the 0.0625 bpw reduction in size is totally worth it. We now have PPL(LLaMA-v2-7B ) = 9.4469 at 1.96 bpw PPL(LLaMA-v2-13B) = 6.8717 at 1.93 bpw PPL(LLaMA-v2-70B) = 4.8568 at 1.85 bpw * iq1_m: scalar dot product * iq1_m: AVX2 dot product * iq1_m: very slightly faster AVX2 dot product * iq1_m: ARM_NEON dot product Works, but very slow (10.5 t/s) * iq1_m: Metal - dequantize works, dot product does not * iq1_m: Metal now works About the same performance as iq1_s. * iq1_m: minor * iq1_m: checking pure iq1_m quantization It is pretty bad: PPL(LLaMA-v2-7B) = 34 if we quantize output.weight with Q4_K. * iiq1_m: slightly faster ARM_NEON dot product 10.5 t/s -> 11.65 t/s * iq1_m: faster ARM_NEON dot product 11.65 t/s -> 14.9 t/s * iq1_m: another minor ARM_NEON dot product improvement 14.9 -> 15.0 t/s * iq1_m: small PPL improvement via super-block scale adjustment After quantizing block scales redo the super-block scale fit. PPL(LLaMA-v2-7B ) = 9.3346 PPL(LLaMA-v2-13B) = 6.8419 PPL(LLaMA-v2-70B) = 4.8294 PPL(Mistral-7B ) = 8.1624 * iq1_m: adapt to CUDA refactoring * iq1_m: remove unused variable We have progressed to warnings being errors. * iq1_m: add to backend-ops tests * iq1_m: fix Windows ARM * iq1_m: use common definition of iq1m_scale_t * cuda: assert -> NO_DEVICE_CODE * iq1_M: PR comments --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-26 14:21:27 +00:00
ftype == LLAMA_FTYPE_MOSTLY_IQ1_S || ftype == LLAMA_FTYPE_MOSTLY_IQ2_S || ftype == LLAMA_FTYPE_MOSTLY_IQ2_M ||
ftype == LLAMA_FTYPE_MOSTLY_IQ1_M) {
quantize: options for output and token embedding tensors qtype (#6239) * quantize: be able to specify the output tensor type * quantize: be able to specify the token embedding tensor type --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-22 18:47:14 +00:00
new_type = GGML_TYPE_Q5_K;
}
else if (new_type != GGML_TYPE_Q8_0) {
new_type = GGML_TYPE_Q6_K;
}
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
}
SOTA 3-bit quants (#5196) * iq3_xxs: quantize/dequantize RMSE seems a bit high-ish at about half-way between q2_K and q3_K, so need to check more. * iq3_xxs: CUDA dequantize works * iq2_xxs: tuning quantization * iq3_xxs: starting to look better PPL on wiki.test.raw LLaMA-v1-7B: 6.4218 LLaMA-v2-7B: 6.3560 Mistral-7B : 6.0717 This is better than Q3_K_XS, with a 5% reduction in quantized model size. * iq3_xxs: CUDA dot product We have PP-512: 5891 t/s TG-128: 143.9 t/s * iq3_xxs: scalar and AVX2 dot products * iq3_xxs: ARM_NEON and Metal Metal performance is decent, ARM_NEON is pathetic * iq3_xxs: slightly better grid points * Faster iq3_xxs and iq2_xs dot products on CUDA * iq3_xxs: add some quant mix * iq3_xxs: fix failing quantization test Dot product still fails. Is this real? * iq3_xxs: hopefully fix ROCm * iq3_xxs: failing tests This time the dot product accuracy did find an actual bug in the AVX2 implementation. * Add IQ3_XXS to test-backend-ops --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-30 13:14:12 +00:00
} else if (name == "token_embd.weight") {
quantize: options for output and token embedding tensors qtype (#6239) * quantize: be able to specify the output tensor type * quantize: be able to specify the token embedding tensor type --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-22 18:47:14 +00:00
if (qs.params->token_embedding_type < GGML_TYPE_COUNT) {
new_type = qs.params->token_embedding_type;
} else {
IQ1_M: 1.75 bpw quantization (#6302) * iq1_m: basics * iq1_m: basics-2 * iq1_m: CUDA dequantize works Very 1st shot I get PPL = 9.76 for LLaMA-v2-7B. * iq1_m: separate shifts for each group of 8 in a block We get PPL(LLaMA-v2-7B ) = 9.2810 PPL(LLaMA-v2-13B) = 6.8105 Not bad, but slightly higher than sqrt(PPL(IQ1_S) * PPL(IQ2_XXS)) which is the expected outcome given that IQ1_M is halfway between IQ1_S and IQ2_XXS in terms of bpw. From this, we would expect PPL = 9.14 for LLaMA-v2-7B PPL = 6.63 for LLaMA-v2-13B * iq1_m: go to 3-bit scales There is slight increase in PPL, but the 0.0625 bpw reduction in size is totally worth it. We now have PPL(LLaMA-v2-7B ) = 9.4469 at 1.96 bpw PPL(LLaMA-v2-13B) = 6.8717 at 1.93 bpw PPL(LLaMA-v2-70B) = 4.8568 at 1.85 bpw * iq1_m: scalar dot product * iq1_m: AVX2 dot product * iq1_m: very slightly faster AVX2 dot product * iq1_m: ARM_NEON dot product Works, but very slow (10.5 t/s) * iq1_m: Metal - dequantize works, dot product does not * iq1_m: Metal now works About the same performance as iq1_s. * iq1_m: minor * iq1_m: checking pure iq1_m quantization It is pretty bad: PPL(LLaMA-v2-7B) = 34 if we quantize output.weight with Q4_K. * iiq1_m: slightly faster ARM_NEON dot product 10.5 t/s -> 11.65 t/s * iq1_m: faster ARM_NEON dot product 11.65 t/s -> 14.9 t/s * iq1_m: another minor ARM_NEON dot product improvement 14.9 -> 15.0 t/s * iq1_m: small PPL improvement via super-block scale adjustment After quantizing block scales redo the super-block scale fit. PPL(LLaMA-v2-7B ) = 9.3346 PPL(LLaMA-v2-13B) = 6.8419 PPL(LLaMA-v2-70B) = 4.8294 PPL(Mistral-7B ) = 8.1624 * iq1_m: adapt to CUDA refactoring * iq1_m: remove unused variable We have progressed to warnings being errors. * iq1_m: add to backend-ops tests * iq1_m: fix Windows ARM * iq1_m: use common definition of iq1m_scale_t * cuda: assert -> NO_DEVICE_CODE * iq1_M: PR comments --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-26 14:21:27 +00:00
if (ftype == LLAMA_FTYPE_MOSTLY_IQ2_XXS || ftype == LLAMA_FTYPE_MOSTLY_IQ2_XS ||
ftype == LLAMA_FTYPE_MOSTLY_IQ1_S || ftype == LLAMA_FTYPE_MOSTLY_IQ1_M) {
quantize: options for output and token embedding tensors qtype (#6239) * quantize: be able to specify the output tensor type * quantize: be able to specify the token embedding tensor type --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-22 18:47:14 +00:00
new_type = GGML_TYPE_Q2_K;
}
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ2_S || ftype == LLAMA_FTYPE_MOSTLY_IQ2_M) {
new_type = GGML_TYPE_IQ3_S;
}
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XXS) {
new_type = GGML_TYPE_IQ3_S;
}
SOTA 3-bit quants (#5196) * iq3_xxs: quantize/dequantize RMSE seems a bit high-ish at about half-way between q2_K and q3_K, so need to check more. * iq3_xxs: CUDA dequantize works * iq2_xxs: tuning quantization * iq3_xxs: starting to look better PPL on wiki.test.raw LLaMA-v1-7B: 6.4218 LLaMA-v2-7B: 6.3560 Mistral-7B : 6.0717 This is better than Q3_K_XS, with a 5% reduction in quantized model size. * iq3_xxs: CUDA dot product We have PP-512: 5891 t/s TG-128: 143.9 t/s * iq3_xxs: scalar and AVX2 dot products * iq3_xxs: ARM_NEON and Metal Metal performance is decent, ARM_NEON is pathetic * iq3_xxs: slightly better grid points * Faster iq3_xxs and iq2_xs dot products on CUDA * iq3_xxs: add some quant mix * iq3_xxs: fix failing quantization test Dot product still fails. Is this real? * iq3_xxs: hopefully fix ROCm * iq3_xxs: failing tests This time the dot product accuracy did find an actual bug in the AVX2 implementation. * Add IQ3_XXS to test-backend-ops --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-30 13:14:12 +00:00
}
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
} else if (ftype == LLAMA_FTYPE_MOSTLY_IQ2_XXS || ftype == LLAMA_FTYPE_MOSTLY_IQ2_XS || ftype == LLAMA_FTYPE_MOSTLY_IQ1_S ||
IQ1_M: 1.75 bpw quantization (#6302) * iq1_m: basics * iq1_m: basics-2 * iq1_m: CUDA dequantize works Very 1st shot I get PPL = 9.76 for LLaMA-v2-7B. * iq1_m: separate shifts for each group of 8 in a block We get PPL(LLaMA-v2-7B ) = 9.2810 PPL(LLaMA-v2-13B) = 6.8105 Not bad, but slightly higher than sqrt(PPL(IQ1_S) * PPL(IQ2_XXS)) which is the expected outcome given that IQ1_M is halfway between IQ1_S and IQ2_XXS in terms of bpw. From this, we would expect PPL = 9.14 for LLaMA-v2-7B PPL = 6.63 for LLaMA-v2-13B * iq1_m: go to 3-bit scales There is slight increase in PPL, but the 0.0625 bpw reduction in size is totally worth it. We now have PPL(LLaMA-v2-7B ) = 9.4469 at 1.96 bpw PPL(LLaMA-v2-13B) = 6.8717 at 1.93 bpw PPL(LLaMA-v2-70B) = 4.8568 at 1.85 bpw * iq1_m: scalar dot product * iq1_m: AVX2 dot product * iq1_m: very slightly faster AVX2 dot product * iq1_m: ARM_NEON dot product Works, but very slow (10.5 t/s) * iq1_m: Metal - dequantize works, dot product does not * iq1_m: Metal now works About the same performance as iq1_s. * iq1_m: minor * iq1_m: checking pure iq1_m quantization It is pretty bad: PPL(LLaMA-v2-7B) = 34 if we quantize output.weight with Q4_K. * iiq1_m: slightly faster ARM_NEON dot product 10.5 t/s -> 11.65 t/s * iq1_m: faster ARM_NEON dot product 11.65 t/s -> 14.9 t/s * iq1_m: another minor ARM_NEON dot product improvement 14.9 -> 15.0 t/s * iq1_m: small PPL improvement via super-block scale adjustment After quantizing block scales redo the super-block scale fit. PPL(LLaMA-v2-7B ) = 9.3346 PPL(LLaMA-v2-13B) = 6.8419 PPL(LLaMA-v2-70B) = 4.8294 PPL(Mistral-7B ) = 8.1624 * iq1_m: adapt to CUDA refactoring * iq1_m: remove unused variable We have progressed to warnings being errors. * iq1_m: add to backend-ops tests * iq1_m: fix Windows ARM * iq1_m: use common definition of iq1m_scale_t * cuda: assert -> NO_DEVICE_CODE * iq1_M: PR comments --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-26 14:21:27 +00:00
ftype == LLAMA_FTYPE_MOSTLY_IQ2_S || ftype == LLAMA_FTYPE_MOSTLY_IQ2_M || ftype == LLAMA_FTYPE_MOSTLY_IQ1_M) {
2-bit quantizations (#4897) * imatrix: load * imatrix: WIP * imatrix: Add Q2_K quantization * imatrix: also guard against Q2_K_S quantization without importance matrix * imatrix: guard even more against low-bit quantization misuse --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:45:56 +00:00
if (name.find("attn_v.weight") != std::string::npos) {
if (qs.model.hparams.n_gqa() >= 4 || qs.model.hparams.n_expert >= 4) new_type = GGML_TYPE_Q4_K;
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
else new_type = ftype == LLAMA_FTYPE_MOSTLY_IQ2_S || ftype == LLAMA_FTYPE_MOSTLY_IQ2_M ? GGML_TYPE_IQ3_S : GGML_TYPE_Q2_K;
2-bit quantizations (#4897) * imatrix: load * imatrix: WIP * imatrix: Add Q2_K quantization * imatrix: also guard against Q2_K_S quantization without importance matrix * imatrix: guard even more against low-bit quantization misuse --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:45:56 +00:00
++qs.i_attention_wv;
}
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
else if (qs.model.hparams.n_expert == 8 && name.find("attn_k.weight") != std::string::npos) {
new_type = GGML_TYPE_Q4_K;
}
2-bit quantizations (#4897) * imatrix: load * imatrix: WIP * imatrix: Add Q2_K quantization * imatrix: also guard against Q2_K_S quantization without importance matrix * imatrix: guard even more against low-bit quantization misuse --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:45:56 +00:00
else if (name.find("ffn_down") != std::string::npos) {
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
if (qs.i_ffn_down < qs.n_ffn_down/8) {
new_type = ftype == LLAMA_FTYPE_MOSTLY_IQ2_S || ftype == LLAMA_FTYPE_MOSTLY_IQ2_M ? GGML_TYPE_IQ3_S : GGML_TYPE_Q2_K;
}
llama : add Q3_K_XS (#5060) * Add Q3_K_XS - intermediate size between Q2_K and Q3_K_S * Q3_K_XS: quanize first 1/8 of ffn_down layers with Q4_K Together with an importance matrix, this brings perplexity for LLaMA-v2-70B below the perplexity of the former Q2_K with a 800 MB smaller quantized model size. --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-22 10:43:33 +00:00
++qs.i_ffn_down;
2-bit quantizations (#4897) * imatrix: load * imatrix: WIP * imatrix: Add Q2_K quantization * imatrix: also guard against Q2_K_S quantization without importance matrix * imatrix: guard even more against low-bit quantization misuse --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:45:56 +00:00
}
1.5 bit quantization (#5453) * iq1_s: WIP basics * iq1_s: CUDA is working * iq1_s: scalar CPU dot product * iq1_s: WIP AVX2 dot product - something is not right * Fix tests * Fix shadow warnings * Fix after merge with latest master * iq1_s: AVX2 finally works * iq1_s: ARM_NEON dot product. Works, but not very fast * iq1_s: better grid * iq1_s: use IQ2_XXS for attn_output At a cost of 0.04 extra bpw this gives a big improvement in PPL. * iq1_s: Metal basics Dequantize works, but not dot product * iq1_s: Metal works, but quite slow As usual, Apple Silicon does not like the code I write. * iq1_s: Tests * iq1_s: slightly faster dot product --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-18 16:16:55 +00:00
else if (name.find("attn_output.weight") != std::string::npos) {
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
if (qs.model.hparams.n_expert == 8) {
new_type = GGML_TYPE_Q5_K;
} else {
IQ1_M: 1.75 bpw quantization (#6302) * iq1_m: basics * iq1_m: basics-2 * iq1_m: CUDA dequantize works Very 1st shot I get PPL = 9.76 for LLaMA-v2-7B. * iq1_m: separate shifts for each group of 8 in a block We get PPL(LLaMA-v2-7B ) = 9.2810 PPL(LLaMA-v2-13B) = 6.8105 Not bad, but slightly higher than sqrt(PPL(IQ1_S) * PPL(IQ2_XXS)) which is the expected outcome given that IQ1_M is halfway between IQ1_S and IQ2_XXS in terms of bpw. From this, we would expect PPL = 9.14 for LLaMA-v2-7B PPL = 6.63 for LLaMA-v2-13B * iq1_m: go to 3-bit scales There is slight increase in PPL, but the 0.0625 bpw reduction in size is totally worth it. We now have PPL(LLaMA-v2-7B ) = 9.4469 at 1.96 bpw PPL(LLaMA-v2-13B) = 6.8717 at 1.93 bpw PPL(LLaMA-v2-70B) = 4.8568 at 1.85 bpw * iq1_m: scalar dot product * iq1_m: AVX2 dot product * iq1_m: very slightly faster AVX2 dot product * iq1_m: ARM_NEON dot product Works, but very slow (10.5 t/s) * iq1_m: Metal - dequantize works, dot product does not * iq1_m: Metal now works About the same performance as iq1_s. * iq1_m: minor * iq1_m: checking pure iq1_m quantization It is pretty bad: PPL(LLaMA-v2-7B) = 34 if we quantize output.weight with Q4_K. * iiq1_m: slightly faster ARM_NEON dot product 10.5 t/s -> 11.65 t/s * iq1_m: faster ARM_NEON dot product 11.65 t/s -> 14.9 t/s * iq1_m: another minor ARM_NEON dot product improvement 14.9 -> 15.0 t/s * iq1_m: small PPL improvement via super-block scale adjustment After quantizing block scales redo the super-block scale fit. PPL(LLaMA-v2-7B ) = 9.3346 PPL(LLaMA-v2-13B) = 6.8419 PPL(LLaMA-v2-70B) = 4.8294 PPL(Mistral-7B ) = 8.1624 * iq1_m: adapt to CUDA refactoring * iq1_m: remove unused variable We have progressed to warnings being errors. * iq1_m: add to backend-ops tests * iq1_m: fix Windows ARM * iq1_m: use common definition of iq1m_scale_t * cuda: assert -> NO_DEVICE_CODE * iq1_M: PR comments --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-26 14:21:27 +00:00
if (ftype == LLAMA_FTYPE_MOSTLY_IQ1_S || ftype == LLAMA_FTYPE_MOSTLY_IQ1_M) new_type = GGML_TYPE_IQ2_XXS;
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ2_S || ftype == LLAMA_FTYPE_MOSTLY_IQ2_M) new_type = GGML_TYPE_IQ3_S;
}
1.5 bit quantization (#5453) * iq1_s: WIP basics * iq1_s: CUDA is working * iq1_s: scalar CPU dot product * iq1_s: WIP AVX2 dot product - something is not right * Fix tests * Fix shadow warnings * Fix after merge with latest master * iq1_s: AVX2 finally works * iq1_s: ARM_NEON dot product. Works, but not very fast * iq1_s: better grid * iq1_s: use IQ2_XXS for attn_output At a cost of 0.04 extra bpw this gives a big improvement in PPL. * iq1_s: Metal basics Dequantize works, but not dot product * iq1_s: Metal works, but quite slow As usual, Apple Silicon does not like the code I write. * iq1_s: Tests * iq1_s: slightly faster dot product --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-18 16:16:55 +00:00
}
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
} else if (name.find("attn_v.weight") != std::string::npos) {
llama : use Q4_K for attn_v for Q2_K_S when n_gqa >= 4 (#4996) Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-17 10:36:37 +00:00
if (ftype == LLAMA_FTYPE_MOSTLY_Q2_K) {
new_type = qs.model.hparams.n_gqa() >= 4 ? GGML_TYPE_Q4_K : GGML_TYPE_Q3_K;
}
else if (ftype == LLAMA_FTYPE_MOSTLY_Q2_K_S && qs.model.hparams.n_gqa() >= 4) {
new_type = GGML_TYPE_Q4_K;
}
iq3_xxs: quards for the no-imatrix situation (#5334) Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-05 10:32:27 +00:00
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XXS) {
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
new_type = qs.model.hparams.n_gqa() >= 4 ? GGML_TYPE_Q4_K : !qs.has_imatrix ? GGML_TYPE_IQ3_S : GGML_TYPE_IQ3_XXS;
}
llama : correction of the attn.v.weight quantization for IQ3_XS (#6209) IQ3_XS was not mentioned, IQ3_S and IQ3_M were present twice. That PR corrects this in the manner which was probably intended initially.
2024-03-22 13:32:02 +00:00
else if ((ftype == LLAMA_FTYPE_MOSTLY_IQ3_XS || ftype == LLAMA_FTYPE_MOSTLY_IQ3_S) && qs.model.hparams.n_gqa() >= 4) {
IQ3_S: a much better alternative to Q3_K (#5676) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * Resurrecting iq3_xs After all the experimentation, nothing was better than this. * Minor PPL improvement via a block scale fudge factor * Minor improvement via 3 neighbours * iq3_xs: working scalar and AVX2 dot products * iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s) * iq3_xs: working Metal implementation * Adding IQ3_M - IQ3_XS mix with mostly Q4_K * iiq3_xs: a 3.4375 bpw variant * iq3_xs: make CUDA work for new version * iq3_xs: make scalar and AVX2 work for new version * iq3_s: make ARM_NEON work with new version * iq3_xs: make new version work on metal Performance is very similar to Q3_K_S * iq3_xs: tiny Metal speed improvement * iq3_xs: tiny Metal speed improvement * Fix stupid warning * Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS * iq3_xs: rename to iq3_s * iq3_s: make tests pass * Move Q3_K_XS mix to 3.25 bpw * Attempt to fix failing tests * Another attempt to fix the Windows builds * Attempt to fix ROCm * ROCm again * iq3_s: partial fix for QK_K = 64 * iq3_s: make it work on metal for QK_K = 64 Pleasent surprise: the coding was super-block size independent, so all it took was to delete some QK_K == 256 guards. * Will this fix ROCm? --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-24 14:23:52 +00:00
new_type = GGML_TYPE_Q4_K;
}
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_M) {
new_type = GGML_TYPE_Q4_K;
}
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M) {
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
new_type = qs.i_attention_wv < 2 ? GGML_TYPE_Q5_K : GGML_TYPE_Q4_K;
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
}
else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_L) new_type = GGML_TYPE_Q5_K;
IQ4_XS: a 4.25 bpw quantization (#5747) * Try IQ4_NL with blocks of 64 - does not look good * iq4_xs: go to super-blocks of 256 and 6-bit scales for blocks of 32 * iq4_xs: CUDA works - 133.2 t/s * iq4_xs: AVX2 dot product * iq4_xs: ARM_NEON dot product * iq4_nl: Metal implementation As usual, Metal / Apple Silicon don't like my quants. * iq3_xs: minor fix * iq4_xs: shrink by using IQ3_S for attn_k and attn_q * iq4_xs: revert using IQ3_S for attn_k and attn_v PPL vs size is good, but CPU performance suffers: on M2 Max TG-128 drops to 21.7 t/s from 28.8, and on a Ryzen-7950X to 14.5 t/s from 15.8 t/s. On CUDA we have 135 t/s when using IQ3_S vs 133 t/s with pure IQ4_XS. * Fix CI * iq4_xs: Added forgotten check for 256 divisibility --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-27 14:34:24 +00:00
else if ((ftype == LLAMA_FTYPE_MOSTLY_IQ4_NL || ftype == LLAMA_FTYPE_MOSTLY_IQ4_XS) && qs.model.hparams.n_gqa() >= 4) {
IQ4_NL: 4-bit non-linear quants with blocks of 32 (#5590) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * iq4_nl: Fix after merging with master * iq4_nl: another fix after merging with master * Use IQ4_NL instead of Q4_K when using k-quants is not possible * Fix typo that makes several tests fail * It was the ggml_vdotq thing missed inside the brackets --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-21 09:39:52 +00:00
new_type = GGML_TYPE_Q5_K;
}
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
else if ((ftype == LLAMA_FTYPE_MOSTLY_Q4_K_M || ftype == LLAMA_FTYPE_MOSTLY_Q5_K_M) &&
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
use_more_bits(qs.i_attention_wv, qs.n_attention_wv)) new_type = GGML_TYPE_Q6_K;
else if (ftype == LLAMA_FTYPE_MOSTLY_Q4_K_S && qs.i_attention_wv < 4) new_type = GGML_TYPE_Q5_K;
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
else if (QK_K == 64 && (ftype == LLAMA_FTYPE_MOSTLY_Q4_K_S || ftype == LLAMA_FTYPE_MOSTLY_Q3_K_S) &&
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
(qs.i_attention_wv < qs.n_attention_wv/8 || qs.i_attention_wv >= 7*qs.n_attention_wv/8)) new_type = GGML_TYPE_Q6_K;
if (qs.model.type == MODEL_70B) {
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
// In the 70B model we have 8 heads sharing the same attn_v weights. As a result, the attn_v.weight tensor is
// 8x smaller compared to attn_q.weight. Hence, we can get a nice boost in quantization accuracy with
// nearly negligible increase in model size by quantizing this tensor with more bits:
if (new_type == GGML_TYPE_Q3_K || new_type == GGML_TYPE_Q4_K) new_type = GGML_TYPE_Q5_K;
}
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
if (qs.model.hparams.n_expert == 8) {
// for the 8-expert model, bumping this to Q8_0 trades just ~128MB
// TODO: explore better strategies
new_type = GGML_TYPE_Q8_0;
}
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
++qs.i_attention_wv;
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
} else if (name.find("attn_k.weight") != std::string::npos) {
if (qs.model.hparams.n_expert == 8) {
// for the 8-expert model, bumping this to Q8_0 trades just ~128MB
// TODO: explore better strategies
new_type = GGML_TYPE_Q8_0;
}
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XS) {
IQ3_S: a much better alternative to Q3_K (#5676) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * Resurrecting iq3_xs After all the experimentation, nothing was better than this. * Minor PPL improvement via a block scale fudge factor * Minor improvement via 3 neighbours * iq3_xs: working scalar and AVX2 dot products * iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s) * iq3_xs: working Metal implementation * Adding IQ3_M - IQ3_XS mix with mostly Q4_K * iiq3_xs: a 3.4375 bpw variant * iq3_xs: make CUDA work for new version * iq3_xs: make scalar and AVX2 work for new version * iq3_s: make ARM_NEON work with new version * iq3_xs: make new version work on metal Performance is very similar to Q3_K_S * iq3_xs: tiny Metal speed improvement * iq3_xs: tiny Metal speed improvement * Fix stupid warning * Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS * iq3_xs: rename to iq3_s * iq3_s: make tests pass * Move Q3_K_XS mix to 3.25 bpw * Attempt to fix failing tests * Another attempt to fix the Windows builds * Attempt to fix ROCm * ROCm again * iq3_s: partial fix for QK_K = 64 * iq3_s: make it work on metal for QK_K = 64 Pleasent surprise: the coding was super-block size independent, so all it took was to delete some QK_K == 256 guards. * Will this fix ROCm? --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-24 14:23:52 +00:00
new_type = GGML_TYPE_IQ3_XXS;
}
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XXS) {
new_type = GGML_TYPE_IQ2_S;
}
IQ3_S: a much better alternative to Q3_K (#5676) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * Resurrecting iq3_xs After all the experimentation, nothing was better than this. * Minor PPL improvement via a block scale fudge factor * Minor improvement via 3 neighbours * iq3_xs: working scalar and AVX2 dot products * iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s) * iq3_xs: working Metal implementation * Adding IQ3_M - IQ3_XS mix with mostly Q4_K * iiq3_xs: a 3.4375 bpw variant * iq3_xs: make CUDA work for new version * iq3_xs: make scalar and AVX2 work for new version * iq3_s: make ARM_NEON work with new version * iq3_xs: make new version work on metal Performance is very similar to Q3_K_S * iq3_xs: tiny Metal speed improvement * iq3_xs: tiny Metal speed improvement * Fix stupid warning * Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS * iq3_xs: rename to iq3_s * iq3_s: make tests pass * Move Q3_K_XS mix to 3.25 bpw * Attempt to fix failing tests * Another attempt to fix the Windows builds * Attempt to fix ROCm * ROCm again * iq3_s: partial fix for QK_K = 64 * iq3_s: make it work on metal for QK_K = 64 Pleasent surprise: the coding was super-block size independent, so all it took was to delete some QK_K == 256 guards. * Will this fix ROCm? --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-24 14:23:52 +00:00
} else if (name.find("attn_q.weight") != std::string::npos) {
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XS) {
IQ3_S: a much better alternative to Q3_K (#5676) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * Resurrecting iq3_xs After all the experimentation, nothing was better than this. * Minor PPL improvement via a block scale fudge factor * Minor improvement via 3 neighbours * iq3_xs: working scalar and AVX2 dot products * iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s) * iq3_xs: working Metal implementation * Adding IQ3_M - IQ3_XS mix with mostly Q4_K * iiq3_xs: a 3.4375 bpw variant * iq3_xs: make CUDA work for new version * iq3_xs: make scalar and AVX2 work for new version * iq3_s: make ARM_NEON work with new version * iq3_xs: make new version work on metal Performance is very similar to Q3_K_S * iq3_xs: tiny Metal speed improvement * iq3_xs: tiny Metal speed improvement * Fix stupid warning * Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS * iq3_xs: rename to iq3_s * iq3_s: make tests pass * Move Q3_K_XS mix to 3.25 bpw * Attempt to fix failing tests * Another attempt to fix the Windows builds * Attempt to fix ROCm * ROCm again * iq3_s: partial fix for QK_K = 64 * iq3_s: make it work on metal for QK_K = 64 Pleasent surprise: the coding was super-block size independent, so all it took was to delete some QK_K == 256 guards. * Will this fix ROCm? --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-24 14:23:52 +00:00
new_type = GGML_TYPE_IQ3_XXS;
llama : add Q3_K_XS (#5060) * Add Q3_K_XS - intermediate size between Q2_K and Q3_K_S * Q3_K_XS: quanize first 1/8 of ffn_down layers with Q4_K Together with an importance matrix, this brings perplexity for LLaMA-v2-70B below the perplexity of the former Q2_K with a 800 MB smaller quantized model size. --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-22 10:43:33 +00:00
}
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XXS) {
new_type = GGML_TYPE_IQ2_S;
}
llama : restore intended k-quants mixes for MoE models (#4872) * Restore intended k-quants quantization mixes for MoE models * Update Q2_K_S values in the quantize tool Still using LLaMA-v1 PPL values in the quant description today does not make much sense. But let's leave this update for another PR. --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-11 19:43:15 +00:00
} else if (name.find("ffn_down") != std::string::npos) {
Fix Q3_K_XS for MoE models (#5113) Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-25 15:58:53 +00:00
auto info = layer_info(qs.i_ffn_down, qs.n_ffn_down, name.c_str());
int i_layer = info.first, n_layer = info.second;
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
if (ftype == LLAMA_FTYPE_MOSTLY_Q2_K) new_type = GGML_TYPE_Q3_K;
IQ3_S: a much better alternative to Q3_K (#5676) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * Resurrecting iq3_xs After all the experimentation, nothing was better than this. * Minor PPL improvement via a block scale fudge factor * Minor improvement via 3 neighbours * iq3_xs: working scalar and AVX2 dot products * iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s) * iq3_xs: working Metal implementation * Adding IQ3_M - IQ3_XS mix with mostly Q4_K * iiq3_xs: a 3.4375 bpw variant * iq3_xs: make CUDA work for new version * iq3_xs: make scalar and AVX2 work for new version * iq3_s: make ARM_NEON work with new version * iq3_xs: make new version work on metal Performance is very similar to Q3_K_S * iq3_xs: tiny Metal speed improvement * iq3_xs: tiny Metal speed improvement * Fix stupid warning * Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS * iq3_xs: rename to iq3_s * iq3_s: make tests pass * Move Q3_K_XS mix to 3.25 bpw * Attempt to fix failing tests * Another attempt to fix the Windows builds * Attempt to fix ROCm * ROCm again * iq3_s: partial fix for QK_K = 64 * iq3_s: make it work on metal for QK_K = 64 Pleasent surprise: the coding was super-block size independent, so all it took was to delete some QK_K == 256 guards. * Will this fix ROCm? --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-24 14:23:52 +00:00
else if (ftype == LLAMA_FTYPE_MOSTLY_Q2_K_S) {
Fix ffn_down quantization mix for MoE models (#4927) * Fix ffn_down quantization mix for MoE models In #4872 I did not consider the part where every third tensor is quantized with more bits. Fir MoE this leads to tensors of the same layer being quantized with different number of bits, which is not considered as a possibility in the inference implementation (it is assumed all experts use the same quantization). * Fix the fix * Review suggestion --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 08:53:39 +00:00
if (i_layer < n_layer/8) new_type = GGML_TYPE_Q4_K;
llama : restore intended k-quants mixes for MoE models (#4872) * Restore intended k-quants quantization mixes for MoE models * Update Q2_K_S values in the quantize tool Still using LLaMA-v1 PPL values in the quant description today does not make much sense. But let's leave this update for another PR. --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-11 19:43:15 +00:00
}
iq3_xxs: quards for the no-imatrix situation (#5334) Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-05 10:32:27 +00:00
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XXS && !qs.has_imatrix) {
new_type = i_layer < n_layer/8 ? GGML_TYPE_Q4_K : GGML_TYPE_Q3_K;
}
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M) {
Fix ffn_down quantization mix for MoE models (#4927) * Fix ffn_down quantization mix for MoE models In #4872 I did not consider the part where every third tensor is quantized with more bits. Fir MoE this leads to tensors of the same layer being quantized with different number of bits, which is not considered as a possibility in the inference implementation (it is assumed all experts use the same quantization). * Fix the fix * Review suggestion --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 08:53:39 +00:00
new_type = i_layer < n_layer/16 ? GGML_TYPE_Q5_K
: arch != LLM_ARCH_FALCON || use_more_bits(i_layer, n_layer) ? GGML_TYPE_Q4_K
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
: GGML_TYPE_Q3_K;
}
IQ3_S: a much better alternative to Q3_K (#5676) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * Resurrecting iq3_xs After all the experimentation, nothing was better than this. * Minor PPL improvement via a block scale fudge factor * Minor improvement via 3 neighbours * iq3_xs: working scalar and AVX2 dot products * iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s) * iq3_xs: working Metal implementation * Adding IQ3_M - IQ3_XS mix with mostly Q4_K * iiq3_xs: a 3.4375 bpw variant * iq3_xs: make CUDA work for new version * iq3_xs: make scalar and AVX2 work for new version * iq3_s: make ARM_NEON work with new version * iq3_xs: make new version work on metal Performance is very similar to Q3_K_S * iq3_xs: tiny Metal speed improvement * iq3_xs: tiny Metal speed improvement * Fix stupid warning * Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS * iq3_xs: rename to iq3_s * iq3_s: make tests pass * Move Q3_K_XS mix to 3.25 bpw * Attempt to fix failing tests * Another attempt to fix the Windows builds * Attempt to fix ROCm * ROCm again * iq3_s: partial fix for QK_K = 64 * iq3_s: make it work on metal for QK_K = 64 Pleasent surprise: the coding was super-block size independent, so all it took was to delete some QK_K == 256 guards. * Will this fix ROCm? --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-24 14:23:52 +00:00
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_M && (i_layer < n_layer/8 ||
(qs.model.hparams.n_expert == 8 && use_more_bits(i_layer, n_layer)))) {
new_type = GGML_TYPE_Q4_K;
}
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_L) {
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
new_type = arch == LLM_ARCH_FALCON ? GGML_TYPE_Q4_K : GGML_TYPE_Q5_K;
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
}
else if (ftype == LLAMA_FTYPE_MOSTLY_Q4_K_M) {
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
if (arch == LLM_ARCH_FALCON) {
Fix ffn_down quantization mix for MoE models (#4927) * Fix ffn_down quantization mix for MoE models In #4872 I did not consider the part where every third tensor is quantized with more bits. Fir MoE this leads to tensors of the same layer being quantized with different number of bits, which is not considered as a possibility in the inference implementation (it is assumed all experts use the same quantization). * Fix the fix * Review suggestion --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 08:53:39 +00:00
new_type = i_layer < n_layer/16 ? GGML_TYPE_Q6_K :
use_more_bits(i_layer, n_layer) ? GGML_TYPE_Q5_K : GGML_TYPE_Q4_K;
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
} else {
Fix ffn_down quantization mix for MoE models (#4927) * Fix ffn_down quantization mix for MoE models In #4872 I did not consider the part where every third tensor is quantized with more bits. Fir MoE this leads to tensors of the same layer being quantized with different number of bits, which is not considered as a possibility in the inference implementation (it is assumed all experts use the same quantization). * Fix the fix * Review suggestion --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 08:53:39 +00:00
if (use_more_bits(i_layer, n_layer)) new_type = GGML_TYPE_Q6_K;
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
}
}
IQ4_XS: a 4.25 bpw quantization (#5747) * Try IQ4_NL with blocks of 64 - does not look good * iq4_xs: go to super-blocks of 256 and 6-bit scales for blocks of 32 * iq4_xs: CUDA works - 133.2 t/s * iq4_xs: AVX2 dot product * iq4_xs: ARM_NEON dot product * iq4_nl: Metal implementation As usual, Metal / Apple Silicon don't like my quants. * iq3_xs: minor fix * iq4_xs: shrink by using IQ3_S for attn_k and attn_q * iq4_xs: revert using IQ3_S for attn_k and attn_v PPL vs size is good, but CPU performance suffers: on M2 Max TG-128 drops to 21.7 t/s from 28.8, and on a Ryzen-7950X to 14.5 t/s from 15.8 t/s. On CUDA we have 135 t/s when using IQ3_S vs 133 t/s with pure IQ4_XS. * Fix CI * iq4_xs: Added forgotten check for 256 divisibility --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-27 14:34:24 +00:00
else if (i_layer < n_layer/8 && (ftype == LLAMA_FTYPE_MOSTLY_IQ4_NL || ftype == LLAMA_FTYPE_MOSTLY_IQ4_XS) && !qs.has_imatrix) {
new_type = GGML_TYPE_Q5_K;
IQ4_NL: 4-bit non-linear quants with blocks of 32 (#5590) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * iq4_nl: Fix after merging with master * iq4_nl: another fix after merging with master * Use IQ4_NL instead of Q4_K when using k-quants is not possible * Fix typo that makes several tests fail * It was the ggml_vdotq thing missed inside the brackets --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-21 09:39:52 +00:00
}
Fix ffn_down quantization mix for MoE models (#4927) * Fix ffn_down quantization mix for MoE models In #4872 I did not consider the part where every third tensor is quantized with more bits. Fir MoE this leads to tensors of the same layer being quantized with different number of bits, which is not considered as a possibility in the inference implementation (it is assumed all experts use the same quantization). * Fix the fix * Review suggestion --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 08:53:39 +00:00
else if (ftype == LLAMA_FTYPE_MOSTLY_Q5_K_M && use_more_bits(i_layer, n_layer)) new_type = GGML_TYPE_Q6_K;
else if (ftype == LLAMA_FTYPE_MOSTLY_Q4_K_S && arch != LLM_ARCH_FALCON && i_layer < n_layer/8) {
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
new_type = GGML_TYPE_Q5_K;
}
ggml : importance matrix support for legacy quants (#4969) * imatrix: adding support for legacy quants * imatrix: guard Q4_0/Q5_0 against ffn_down craziness --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-16 17:51:26 +00:00
else if ((ftype == LLAMA_FTYPE_MOSTLY_Q4_0 || ftype == LLAMA_FTYPE_MOSTLY_Q5_0)
&& qs.has_imatrix && i_layer < n_layer/8) {
// Guard against craziness in the first few ffn_down layers that can happen even with imatrix for Q4_0/Q5_0.
// We only do it when an imatrix is provided because a) we want to make sure that one can always get the
// same quantization as before imatrix stuff, and b) Q4_1/Q5_1 do go crazy on ffn_down without an imatrix.
new_type = ftype == LLAMA_FTYPE_MOSTLY_Q4_0 ? GGML_TYPE_Q4_1 : GGML_TYPE_Q5_1;
}
llama : add Q3_K_XS (#5060) * Add Q3_K_XS - intermediate size between Q2_K and Q3_K_S * Q3_K_XS: quanize first 1/8 of ffn_down layers with Q4_K Together with an importance matrix, this brings perplexity for LLaMA-v2-70B below the perplexity of the former Q2_K with a 800 MB smaller quantized model size. --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-22 10:43:33 +00:00
++qs.i_ffn_down;
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
} else if (name.find("attn_output.weight") != std::string::npos) {
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
if (arch != LLM_ARCH_FALCON) {
Make Q3_K_S be the same as olf Q3_K_L for Mixtral-8x7B (#4906) Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:44:30 +00:00
if (qs.model.hparams.n_expert == 8) {
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
if (ftype == LLAMA_FTYPE_MOSTLY_Q2_K || ftype == LLAMA_FTYPE_MOSTLY_IQ3_XS || ftype == LLAMA_FTYPE_MOSTLY_IQ3_XXS ||
IQ4_NL: 4-bit non-linear quants with blocks of 32 (#5590) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * iq4_nl: Fix after merging with master * iq4_nl: another fix after merging with master * Use IQ4_NL instead of Q4_K when using k-quants is not possible * Fix typo that makes several tests fail * It was the ggml_vdotq thing missed inside the brackets --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-21 09:39:52 +00:00
ftype == LLAMA_FTYPE_MOSTLY_Q3_K_S || ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M || ftype == LLAMA_FTYPE_MOSTLY_IQ4_NL ||
IQ3_S: a much better alternative to Q3_K (#5676) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * Resurrecting iq3_xs After all the experimentation, nothing was better than this. * Minor PPL improvement via a block scale fudge factor * Minor improvement via 3 neighbours * iq3_xs: working scalar and AVX2 dot products * iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s) * iq3_xs: working Metal implementation * Adding IQ3_M - IQ3_XS mix with mostly Q4_K * iiq3_xs: a 3.4375 bpw variant * iq3_xs: make CUDA work for new version * iq3_xs: make scalar and AVX2 work for new version * iq3_s: make ARM_NEON work with new version * iq3_xs: make new version work on metal Performance is very similar to Q3_K_S * iq3_xs: tiny Metal speed improvement * iq3_xs: tiny Metal speed improvement * Fix stupid warning * Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS * iq3_xs: rename to iq3_s * iq3_s: make tests pass * Move Q3_K_XS mix to 3.25 bpw * Attempt to fix failing tests * Another attempt to fix the Windows builds * Attempt to fix ROCm * ROCm again * iq3_s: partial fix for QK_K = 64 * iq3_s: make it work on metal for QK_K = 64 Pleasent surprise: the coding was super-block size independent, so all it took was to delete some QK_K == 256 guards. * Will this fix ROCm? --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-24 14:23:52 +00:00
ftype == LLAMA_FTYPE_MOSTLY_Q4_K_S || ftype == LLAMA_FTYPE_MOSTLY_Q4_K_M || ftype == LLAMA_FTYPE_MOSTLY_IQ3_S ||
IQ4_XS: a 4.25 bpw quantization (#5747) * Try IQ4_NL with blocks of 64 - does not look good * iq4_xs: go to super-blocks of 256 and 6-bit scales for blocks of 32 * iq4_xs: CUDA works - 133.2 t/s * iq4_xs: AVX2 dot product * iq4_xs: ARM_NEON dot product * iq4_nl: Metal implementation As usual, Metal / Apple Silicon don't like my quants. * iq3_xs: minor fix * iq4_xs: shrink by using IQ3_S for attn_k and attn_q * iq4_xs: revert using IQ3_S for attn_k and attn_v PPL vs size is good, but CPU performance suffers: on M2 Max TG-128 drops to 21.7 t/s from 28.8, and on a Ryzen-7950X to 14.5 t/s from 15.8 t/s. On CUDA we have 135 t/s when using IQ3_S vs 133 t/s with pure IQ4_XS. * Fix CI * iq4_xs: Added forgotten check for 256 divisibility --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-27 14:34:24 +00:00
ftype == LLAMA_FTYPE_MOSTLY_IQ3_M || ftype == LLAMA_FTYPE_MOSTLY_IQ4_XS) {
Make Q3_K_S be the same as olf Q3_K_L for Mixtral-8x7B (#4906) Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:44:30 +00:00
new_type = GGML_TYPE_Q5_K;
}
} else {
IQ3_S: a much better alternative to Q3_K (#5676) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * Resurrecting iq3_xs After all the experimentation, nothing was better than this. * Minor PPL improvement via a block scale fudge factor * Minor improvement via 3 neighbours * iq3_xs: working scalar and AVX2 dot products * iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s) * iq3_xs: working Metal implementation * Adding IQ3_M - IQ3_XS mix with mostly Q4_K * iiq3_xs: a 3.4375 bpw variant * iq3_xs: make CUDA work for new version * iq3_xs: make scalar and AVX2 work for new version * iq3_s: make ARM_NEON work with new version * iq3_xs: make new version work on metal Performance is very similar to Q3_K_S * iq3_xs: tiny Metal speed improvement * iq3_xs: tiny Metal speed improvement * Fix stupid warning * Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS * iq3_xs: rename to iq3_s * iq3_s: make tests pass * Move Q3_K_XS mix to 3.25 bpw * Attempt to fix failing tests * Another attempt to fix the Windows builds * Attempt to fix ROCm * ROCm again * iq3_s: partial fix for QK_K = 64 * iq3_s: make it work on metal for QK_K = 64 Pleasent surprise: the coding was super-block size independent, so all it took was to delete some QK_K == 256 guards. * Will this fix ROCm? --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-24 14:23:52 +00:00
if (ftype == LLAMA_FTYPE_MOSTLY_Q2_K ) new_type = GGML_TYPE_Q3_K;
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XXS) new_type = GGML_TYPE_IQ3_S;
IQ3_S: a much better alternative to Q3_K (#5676) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * Resurrecting iq3_xs After all the experimentation, nothing was better than this. * Minor PPL improvement via a block scale fudge factor * Minor improvement via 3 neighbours * iq3_xs: working scalar and AVX2 dot products * iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s) * iq3_xs: working Metal implementation * Adding IQ3_M - IQ3_XS mix with mostly Q4_K * iiq3_xs: a 3.4375 bpw variant * iq3_xs: make CUDA work for new version * iq3_xs: make scalar and AVX2 work for new version * iq3_s: make ARM_NEON work with new version * iq3_xs: make new version work on metal Performance is very similar to Q3_K_S * iq3_xs: tiny Metal speed improvement * iq3_xs: tiny Metal speed improvement * Fix stupid warning * Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS * iq3_xs: rename to iq3_s * iq3_s: make tests pass * Move Q3_K_XS mix to 3.25 bpw * Attempt to fix failing tests * Another attempt to fix the Windows builds * Attempt to fix ROCm * ROCm again * iq3_s: partial fix for QK_K = 64 * iq3_s: make it work on metal for QK_K = 64 Pleasent surprise: the coding was super-block size independent, so all it took was to delete some QK_K == 256 guards. * Will this fix ROCm? --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-24 14:23:52 +00:00
else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M ) new_type = GGML_TYPE_Q4_K;
else if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_L ) new_type = GGML_TYPE_Q5_K;
else if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_M ) new_type = GGML_TYPE_Q4_K;
Make Q3_K_S be the same as olf Q3_K_L for Mixtral-8x7B (#4906) Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:44:30 +00:00
}
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
} else {
if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_L) new_type = GGML_TYPE_Q4_K;
}
}
else if (name.find("attn_qkv.weight") != std::string::npos) {
IQ3_S: a much better alternative to Q3_K (#5676) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * Resurrecting iq3_xs After all the experimentation, nothing was better than this. * Minor PPL improvement via a block scale fudge factor * Minor improvement via 3 neighbours * iq3_xs: working scalar and AVX2 dot products * iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s) * iq3_xs: working Metal implementation * Adding IQ3_M - IQ3_XS mix with mostly Q4_K * iiq3_xs: a 3.4375 bpw variant * iq3_xs: make CUDA work for new version * iq3_xs: make scalar and AVX2 work for new version * iq3_s: make ARM_NEON work with new version * iq3_xs: make new version work on metal Performance is very similar to Q3_K_S * iq3_xs: tiny Metal speed improvement * iq3_xs: tiny Metal speed improvement * Fix stupid warning * Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS * iq3_xs: rename to iq3_s * iq3_s: make tests pass * Move Q3_K_XS mix to 3.25 bpw * Attempt to fix failing tests * Another attempt to fix the Windows builds * Attempt to fix ROCm * ROCm again * iq3_s: partial fix for QK_K = 64 * iq3_s: make it work on metal for QK_K = 64 Pleasent surprise: the coding was super-block size independent, so all it took was to delete some QK_K == 256 guards. * Will this fix ROCm? --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-24 14:23:52 +00:00
if (ftype == LLAMA_FTYPE_MOSTLY_Q3_K_M || ftype == LLAMA_FTYPE_MOSTLY_Q3_K_L || ftype == LLAMA_FTYPE_MOSTLY_IQ3_M) {
new_type = GGML_TYPE_Q4_K;
}
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
else if (ftype == LLAMA_FTYPE_MOSTLY_Q4_K_M) new_type = GGML_TYPE_Q5_K;
else if (ftype == LLAMA_FTYPE_MOSTLY_Q5_K_M) new_type = GGML_TYPE_Q6_K;
}
llama : add Q3_K_XS (#5060) * Add Q3_K_XS - intermediate size between Q2_K and Q3_K_S * Q3_K_XS: quanize first 1/8 of ffn_down layers with Q4_K Together with an importance matrix, this brings perplexity for LLaMA-v2-70B below the perplexity of the former Q2_K with a 800 MB smaller quantized model size. --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-22 10:43:33 +00:00
else if (name.find("ffn_gate") != std::string::npos) {
Fix Q3_K_XS for MoE models (#5113) Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-25 15:58:53 +00:00
auto info = layer_info(qs.i_ffn_gate, qs.n_ffn_gate, name.c_str());
int i_layer = info.first, n_layer = info.second;
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XS && (i_layer >= n_layer/8 && i_layer < 7*n_layer/8)) {
IQ3_S: a much better alternative to Q3_K (#5676) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * Resurrecting iq3_xs After all the experimentation, nothing was better than this. * Minor PPL improvement via a block scale fudge factor * Minor improvement via 3 neighbours * iq3_xs: working scalar and AVX2 dot products * iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s) * iq3_xs: working Metal implementation * Adding IQ3_M - IQ3_XS mix with mostly Q4_K * iiq3_xs: a 3.4375 bpw variant * iq3_xs: make CUDA work for new version * iq3_xs: make scalar and AVX2 work for new version * iq3_s: make ARM_NEON work with new version * iq3_xs: make new version work on metal Performance is very similar to Q3_K_S * iq3_xs: tiny Metal speed improvement * iq3_xs: tiny Metal speed improvement * Fix stupid warning * Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS * iq3_xs: rename to iq3_s * iq3_s: make tests pass * Move Q3_K_XS mix to 3.25 bpw * Attempt to fix failing tests * Another attempt to fix the Windows builds * Attempt to fix ROCm * ROCm again * iq3_s: partial fix for QK_K = 64 * iq3_s: make it work on metal for QK_K = 64 Pleasent surprise: the coding was super-block size independent, so all it took was to delete some QK_K == 256 guards. * Will this fix ROCm? --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-24 14:23:52 +00:00
new_type = GGML_TYPE_IQ3_XXS;
llama : add Q3_K_XS (#5060) * Add Q3_K_XS - intermediate size between Q2_K and Q3_K_S * Q3_K_XS: quanize first 1/8 of ffn_down layers with Q4_K Together with an importance matrix, this brings perplexity for LLaMA-v2-70B below the perplexity of the former Q2_K with a 800 MB smaller quantized model size. --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-22 10:43:33 +00:00
}
++qs.i_ffn_gate;
}
else if (name.find("ffn_up") != std::string::npos) {
Fix Q3_K_XS for MoE models (#5113) Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-25 15:58:53 +00:00
auto info = layer_info(qs.i_ffn_up, qs.n_ffn_up, name.c_str());
int i_layer = info.first, n_layer = info.second;
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
if (ftype == LLAMA_FTYPE_MOSTLY_IQ3_XS && (i_layer >= n_layer/8 && i_layer < 7*n_layer/8)) {
IQ3_S: a much better alternative to Q3_K (#5676) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * Resurrecting iq3_xs After all the experimentation, nothing was better than this. * Minor PPL improvement via a block scale fudge factor * Minor improvement via 3 neighbours * iq3_xs: working scalar and AVX2 dot products * iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s) * iq3_xs: working Metal implementation * Adding IQ3_M - IQ3_XS mix with mostly Q4_K * iiq3_xs: a 3.4375 bpw variant * iq3_xs: make CUDA work for new version * iq3_xs: make scalar and AVX2 work for new version * iq3_s: make ARM_NEON work with new version * iq3_xs: make new version work on metal Performance is very similar to Q3_K_S * iq3_xs: tiny Metal speed improvement * iq3_xs: tiny Metal speed improvement * Fix stupid warning * Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS * iq3_xs: rename to iq3_s * iq3_s: make tests pass * Move Q3_K_XS mix to 3.25 bpw * Attempt to fix failing tests * Another attempt to fix the Windows builds * Attempt to fix ROCm * ROCm again * iq3_s: partial fix for QK_K = 64 * iq3_s: make it work on metal for QK_K = 64 Pleasent surprise: the coding was super-block size independent, so all it took was to delete some QK_K == 256 guards. * Will this fix ROCm? --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-24 14:23:52 +00:00
new_type = GGML_TYPE_IQ3_XXS;
llama : add Q3_K_XS (#5060) * Add Q3_K_XS - intermediate size between Q2_K and Q3_K_S * Q3_K_XS: quanize first 1/8 of ffn_down layers with Q4_K Together with an importance matrix, this brings perplexity for LLaMA-v2-70B below the perplexity of the former Q2_K with a 800 MB smaller quantized model size. --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-22 10:43:33 +00:00
}
++qs.i_ffn_up;
}
llama : fix quantization when tensors are missing (#5423)
2024-02-12 18:14:39 +00:00
llama : add Q3_K_XS (#5060) * Add Q3_K_XS - intermediate size between Q2_K and Q3_K_S * Q3_K_XS: quanize first 1/8 of ffn_down layers with Q4_K Together with an importance matrix, this brings perplexity for LLaMA-v2-70B below the perplexity of the former Q2_K with a 800 MB smaller quantized model size. --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-22 10:43:33 +00:00
// if (ftype == LLAMA_FTYPE_MOSTLY_Q2_K) new_type = GGML_TYPE_Q3_K;
//}
llama : restore intended k-quants mixes for MoE models (#4872) * Restore intended k-quants quantization mixes for MoE models * Update Q2_K_S values in the quantize tool Still using LLaMA-v1 PPL values in the quant description today does not make much sense. But let's leave this update for another PR. --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-11 19:43:15 +00:00
// IK: let's remove this, else Q2_K is almost the same as Q3_K_S
//else if (name.find("ffn_gate") != std::string::npos || name.find("ffn_up") != std::string::npos) {
// if (ftype == LLAMA_FTYPE_MOSTLY_Q2_K) new_type = GGML_TYPE_Q3_K;
//}
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
// This can be used to reduce the size of the Q5_K_S model.
// The associated PPL increase is fully in line with the size reduction
//else {
// if (ftype == LLAMA_FTYPE_MOSTLY_Q5_K_S) new_type = GGML_TYPE_Q4_K;
//}
bool convert_incompatible_tensor = false;
if (new_type == GGML_TYPE_Q2_K || new_type == GGML_TYPE_Q3_K || new_type == GGML_TYPE_Q4_K ||
IQ4_XS: a 4.25 bpw quantization (#5747) * Try IQ4_NL with blocks of 64 - does not look good * iq4_xs: go to super-blocks of 256 and 6-bit scales for blocks of 32 * iq4_xs: CUDA works - 133.2 t/s * iq4_xs: AVX2 dot product * iq4_xs: ARM_NEON dot product * iq4_nl: Metal implementation As usual, Metal / Apple Silicon don't like my quants. * iq3_xs: minor fix * iq4_xs: shrink by using IQ3_S for attn_k and attn_q * iq4_xs: revert using IQ3_S for attn_k and attn_v PPL vs size is good, but CPU performance suffers: on M2 Max TG-128 drops to 21.7 t/s from 28.8, and on a Ryzen-7950X to 14.5 t/s from 15.8 t/s. On CUDA we have 135 t/s when using IQ3_S vs 133 t/s with pure IQ4_XS. * Fix CI * iq4_xs: Added forgotten check for 256 divisibility --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-27 14:34:24 +00:00
new_type == GGML_TYPE_Q5_K || new_type == GGML_TYPE_Q6_K || new_type == GGML_TYPE_IQ4_XS ||
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
new_type == GGML_TYPE_IQ2_XS || new_type == GGML_TYPE_IQ2_XXS || new_type == GGML_TYPE_IQ2_S ||
IQ1_M: 1.75 bpw quantization (#6302) * iq1_m: basics * iq1_m: basics-2 * iq1_m: CUDA dequantize works Very 1st shot I get PPL = 9.76 for LLaMA-v2-7B. * iq1_m: separate shifts for each group of 8 in a block We get PPL(LLaMA-v2-7B ) = 9.2810 PPL(LLaMA-v2-13B) = 6.8105 Not bad, but slightly higher than sqrt(PPL(IQ1_S) * PPL(IQ2_XXS)) which is the expected outcome given that IQ1_M is halfway between IQ1_S and IQ2_XXS in terms of bpw. From this, we would expect PPL = 9.14 for LLaMA-v2-7B PPL = 6.63 for LLaMA-v2-13B * iq1_m: go to 3-bit scales There is slight increase in PPL, but the 0.0625 bpw reduction in size is totally worth it. We now have PPL(LLaMA-v2-7B ) = 9.4469 at 1.96 bpw PPL(LLaMA-v2-13B) = 6.8717 at 1.93 bpw PPL(LLaMA-v2-70B) = 4.8568 at 1.85 bpw * iq1_m: scalar dot product * iq1_m: AVX2 dot product * iq1_m: very slightly faster AVX2 dot product * iq1_m: ARM_NEON dot product Works, but very slow (10.5 t/s) * iq1_m: Metal - dequantize works, dot product does not * iq1_m: Metal now works About the same performance as iq1_s. * iq1_m: minor * iq1_m: checking pure iq1_m quantization It is pretty bad: PPL(LLaMA-v2-7B) = 34 if we quantize output.weight with Q4_K. * iiq1_m: slightly faster ARM_NEON dot product 10.5 t/s -> 11.65 t/s * iq1_m: faster ARM_NEON dot product 11.65 t/s -> 14.9 t/s * iq1_m: another minor ARM_NEON dot product improvement 14.9 -> 15.0 t/s * iq1_m: small PPL improvement via super-block scale adjustment After quantizing block scales redo the super-block scale fit. PPL(LLaMA-v2-7B ) = 9.3346 PPL(LLaMA-v2-13B) = 6.8419 PPL(LLaMA-v2-70B) = 4.8294 PPL(Mistral-7B ) = 8.1624 * iq1_m: adapt to CUDA refactoring * iq1_m: remove unused variable We have progressed to warnings being errors. * iq1_m: add to backend-ops tests * iq1_m: fix Windows ARM * iq1_m: use common definition of iq1m_scale_t * cuda: assert -> NO_DEVICE_CODE * iq1_M: PR comments --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-26 14:21:27 +00:00
new_type == GGML_TYPE_IQ3_XXS || new_type == GGML_TYPE_IQ1_S || new_type == GGML_TYPE_IQ3_S ||
new_type == GGML_TYPE_IQ1_M) {
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
int nx = tensor->ne[0];
int ny = tensor->ne[1];
if (nx % QK_K != 0) {
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
LLAMA_LOG_WARN("\n\n%s : tensor cols %d x %d are not divisible by %d, required for %s", __func__, nx, ny, QK_K, ggml_type_name(new_type));
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
convert_incompatible_tensor = true;
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
} else {
++qs.n_k_quantized;
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
}
}
if (convert_incompatible_tensor) {
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
switch (new_type) {
llama : check for 256 divisibility for IQ2_XS, IQ2_XXS (#4950) Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-15 08:09:38 +00:00
case GGML_TYPE_IQ2_XXS:
case GGML_TYPE_IQ2_XS:
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
case GGML_TYPE_IQ2_S:
SOTA 3-bit quants (#5196) * iq3_xxs: quantize/dequantize RMSE seems a bit high-ish at about half-way between q2_K and q3_K, so need to check more. * iq3_xxs: CUDA dequantize works * iq2_xxs: tuning quantization * iq3_xxs: starting to look better PPL on wiki.test.raw LLaMA-v1-7B: 6.4218 LLaMA-v2-7B: 6.3560 Mistral-7B : 6.0717 This is better than Q3_K_XS, with a 5% reduction in quantized model size. * iq3_xxs: CUDA dot product We have PP-512: 5891 t/s TG-128: 143.9 t/s * iq3_xxs: scalar and AVX2 dot products * iq3_xxs: ARM_NEON and Metal Metal performance is decent, ARM_NEON is pathetic * iq3_xxs: slightly better grid points * Faster iq3_xxs and iq2_xs dot products on CUDA * iq3_xxs: add some quant mix * iq3_xxs: fix failing quantization test Dot product still fails. Is this real? * iq3_xxs: hopefully fix ROCm * iq3_xxs: failing tests This time the dot product accuracy did find an actual bug in the AVX2 implementation. * Add IQ3_XXS to test-backend-ops --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-30 13:14:12 +00:00
case GGML_TYPE_IQ3_XXS:
IQ3_S: a much better alternative to Q3_K (#5676) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * Resurrecting iq3_xs After all the experimentation, nothing was better than this. * Minor PPL improvement via a block scale fudge factor * Minor improvement via 3 neighbours * iq3_xs: working scalar and AVX2 dot products * iq3_xs: ARM_NEON dot product - works but extremely slow (10 t/s) * iq3_xs: working Metal implementation * Adding IQ3_M - IQ3_XS mix with mostly Q4_K * iiq3_xs: a 3.4375 bpw variant * iq3_xs: make CUDA work for new version * iq3_xs: make scalar and AVX2 work for new version * iq3_s: make ARM_NEON work with new version * iq3_xs: make new version work on metal Performance is very similar to Q3_K_S * iq3_xs: tiny Metal speed improvement * iq3_xs: tiny Metal speed improvement * Fix stupid warning * Q3_K_XS now uses a mix of IQ3_XS and IQ3_XXS * iq3_xs: rename to iq3_s * iq3_s: make tests pass * Move Q3_K_XS mix to 3.25 bpw * Attempt to fix failing tests * Another attempt to fix the Windows builds * Attempt to fix ROCm * ROCm again * iq3_s: partial fix for QK_K = 64 * iq3_s: make it work on metal for QK_K = 64 Pleasent surprise: the coding was super-block size independent, so all it took was to delete some QK_K == 256 guards. * Will this fix ROCm? --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-24 14:23:52 +00:00
case GGML_TYPE_IQ3_S:
1.5 bit quantization (#5453) * iq1_s: WIP basics * iq1_s: CUDA is working * iq1_s: scalar CPU dot product * iq1_s: WIP AVX2 dot product - something is not right * Fix tests * Fix shadow warnings * Fix after merge with latest master * iq1_s: AVX2 finally works * iq1_s: ARM_NEON dot product. Works, but not very fast * iq1_s: better grid * iq1_s: use IQ2_XXS for attn_output At a cost of 0.04 extra bpw this gives a big improvement in PPL. * iq1_s: Metal basics Dequantize works, but not dot product * iq1_s: Metal works, but quite slow As usual, Apple Silicon does not like the code I write. * iq1_s: Tests * iq1_s: slightly faster dot product --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-18 16:16:55 +00:00
case GGML_TYPE_IQ1_S:
IQ1_M: 1.75 bpw quantization (#6302) * iq1_m: basics * iq1_m: basics-2 * iq1_m: CUDA dequantize works Very 1st shot I get PPL = 9.76 for LLaMA-v2-7B. * iq1_m: separate shifts for each group of 8 in a block We get PPL(LLaMA-v2-7B ) = 9.2810 PPL(LLaMA-v2-13B) = 6.8105 Not bad, but slightly higher than sqrt(PPL(IQ1_S) * PPL(IQ2_XXS)) which is the expected outcome given that IQ1_M is halfway between IQ1_S and IQ2_XXS in terms of bpw. From this, we would expect PPL = 9.14 for LLaMA-v2-7B PPL = 6.63 for LLaMA-v2-13B * iq1_m: go to 3-bit scales There is slight increase in PPL, but the 0.0625 bpw reduction in size is totally worth it. We now have PPL(LLaMA-v2-7B ) = 9.4469 at 1.96 bpw PPL(LLaMA-v2-13B) = 6.8717 at 1.93 bpw PPL(LLaMA-v2-70B) = 4.8568 at 1.85 bpw * iq1_m: scalar dot product * iq1_m: AVX2 dot product * iq1_m: very slightly faster AVX2 dot product * iq1_m: ARM_NEON dot product Works, but very slow (10.5 t/s) * iq1_m: Metal - dequantize works, dot product does not * iq1_m: Metal now works About the same performance as iq1_s. * iq1_m: minor * iq1_m: checking pure iq1_m quantization It is pretty bad: PPL(LLaMA-v2-7B) = 34 if we quantize output.weight with Q4_K. * iiq1_m: slightly faster ARM_NEON dot product 10.5 t/s -> 11.65 t/s * iq1_m: faster ARM_NEON dot product 11.65 t/s -> 14.9 t/s * iq1_m: another minor ARM_NEON dot product improvement 14.9 -> 15.0 t/s * iq1_m: small PPL improvement via super-block scale adjustment After quantizing block scales redo the super-block scale fit. PPL(LLaMA-v2-7B ) = 9.3346 PPL(LLaMA-v2-13B) = 6.8419 PPL(LLaMA-v2-70B) = 4.8294 PPL(Mistral-7B ) = 8.1624 * iq1_m: adapt to CUDA refactoring * iq1_m: remove unused variable We have progressed to warnings being errors. * iq1_m: add to backend-ops tests * iq1_m: fix Windows ARM * iq1_m: use common definition of iq1m_scale_t * cuda: assert -> NO_DEVICE_CODE * iq1_M: PR comments --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-26 14:21:27 +00:00
case GGML_TYPE_IQ1_M:
IQ4_NL: 4-bit non-linear quants with blocks of 32 (#5590) * iq4_nl: squash commits for easier rebase * Basics (quantize, dequantize) * CUDA dequantize and dot product * Slightly faster CUDA dot product (120 t/s) * Switch to 6-bit scales * Scalar dot product * AVX2 dot product * ARM_NEON dot product * Works on metal, but still slow * Slightly better Metal dot product * Another small Metal improvement * Metal dot product is getting there * Faster CUDA dot product * Add 1/8 ffn_down layers as Q5_K when no imatrix has been provided * Report the actual bpw * Add _xs mix that is 4.05 bpw for non-MoE models * Remove IQ4_XS for now, slightly adjust kvalues_iq4nl * AVX2 dot product uses Q8_0 instead of Q8_K * Add to test-backend-ops * Minor fix * Also use use Q5_K for attn_output in MoE models * Fixes after merging latest master * Switching to blocks of 32 * AVX2 for blocks of 32 * Scaler dot product for blocks of 32 * ARM_NEON dot product for blocks of 32 * Metal kernels for blocks of 32 * Slightly faster Metal kernels * iq4_nl: Fix after merging with master * iq4_nl: another fix after merging with master * Use IQ4_NL instead of Q4_K when using k-quants is not possible * Fix typo that makes several tests fail * It was the ggml_vdotq thing missed inside the brackets --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-21 09:39:52 +00:00
case GGML_TYPE_Q2_K:
IQ4_XS: a 4.25 bpw quantization (#5747) * Try IQ4_NL with blocks of 64 - does not look good * iq4_xs: go to super-blocks of 256 and 6-bit scales for blocks of 32 * iq4_xs: CUDA works - 133.2 t/s * iq4_xs: AVX2 dot product * iq4_xs: ARM_NEON dot product * iq4_nl: Metal implementation As usual, Metal / Apple Silicon don't like my quants. * iq3_xs: minor fix * iq4_xs: shrink by using IQ3_S for attn_k and attn_q * iq4_xs: revert using IQ3_S for attn_k and attn_v PPL vs size is good, but CPU performance suffers: on M2 Max TG-128 drops to 21.7 t/s from 28.8, and on a Ryzen-7950X to 14.5 t/s from 15.8 t/s. On CUDA we have 135 t/s when using IQ3_S vs 133 t/s with pure IQ4_XS. * Fix CI * iq4_xs: Added forgotten check for 256 divisibility --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-27 14:34:24 +00:00
case GGML_TYPE_Q3_K:
case GGML_TYPE_IQ4_XS: new_type = GGML_TYPE_IQ4_NL; break;
case GGML_TYPE_Q4_K: new_type = GGML_TYPE_Q5_0; break;
case GGML_TYPE_Q5_K: new_type = GGML_TYPE_Q5_1; break;
case GGML_TYPE_Q6_K: new_type = GGML_TYPE_Q8_0; break;
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
default: throw std::runtime_error("\nUnsupported tensor size encountered\n");
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
}
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
LLAMA_LOG_WARN(" - using fallback quantization %s\n", ggml_type_name(new_type));
++qs.n_fallback;
llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691) * Add support for quantizing already quantized models * Threaded dequantizing and f16 to f32 conversion * Clean up thread blocks with spares calculation a bit * Use std::runtime_error exceptions.
2023-06-10 07:59:17 +00:00
}
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
return new_type;
llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691) * Add support for quantizing already quantized models * Threaded dequantizing and f16 to f32 conversion * Clean up thread blocks with spares calculation a bit * Use std::runtime_error exceptions.
2023-06-10 07:59:17 +00:00
}
llama : add Command R Plus support (#6491) * Add Command R Plus GGUF * Add Command R Plus GGUF * Loading works up to LayerNorm2D * Export new tensors in 1D so they are not quantized. * Fix embedding layer based on Noeda's example * Whitespace * Add line * Fix unexpected tokens on MPS. Re-add F16 fix. ((Noeda) * dranger003: Fix block index overflow in CUDA dequantizing. * Reverted blocked multiplication code as it still has issues and could affect other Llama arches * export norms as f32 * fix overflow issues during quant and other cleanup * Type convention Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * dranger003: Fix more int overflow during quant. --------- Co-authored-by: S <seast@Ss-Mac-Studio.local> Co-authored-by: S <s@example.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-09 08:16:13 +00:00
static size_t llama_tensor_quantize_internal(enum ggml_type new_type, const float * f32_data, void * new_data, const int64_t chunk_size, int64_t nrows, int64_t n_per_row, const float * imatrix, std::vector<std::thread> & workers, const int nthread) {
llama : refactor internal quantization functions (#5830)
2024-03-02 14:19:09 +00:00
if (nthread < 2) {
// single-thread
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
size_t new_size = ggml_quantize_chunk(new_type, f32_data, new_data, 0, nrows, n_per_row, imatrix);
if (!ggml_validate_row_data(new_type, new_data, new_size)) {
throw std::runtime_error("quantized data validation failed");
}
return new_size;
llama : refactor internal quantization functions (#5830)
2024-03-02 14:19:09 +00:00
}
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
std::mutex mutex;
int64_t counter = 0;
size_t new_size = 0;
bool valid = true;
auto compute = [&mutex, &counter, &new_size, &valid, new_type, f32_data, new_data, chunk_size,
llama : refactor internal quantization functions (#5830)
2024-03-02 14:19:09 +00:00
nrows, n_per_row, imatrix]() {
llama : add Command R Plus support (#6491) * Add Command R Plus GGUF * Add Command R Plus GGUF * Loading works up to LayerNorm2D * Export new tensors in 1D so they are not quantized. * Fix embedding layer based on Noeda's example * Whitespace * Add line * Fix unexpected tokens on MPS. Re-add F16 fix. ((Noeda) * dranger003: Fix block index overflow in CUDA dequantizing. * Reverted blocked multiplication code as it still has issues and could affect other Llama arches * export norms as f32 * fix overflow issues during quant and other cleanup * Type convention Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * dranger003: Fix more int overflow during quant. --------- Co-authored-by: S <seast@Ss-Mac-Studio.local> Co-authored-by: S <s@example.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-09 08:16:13 +00:00
const int64_t nrows_per_chunk = chunk_size / n_per_row;
llama : refactor internal quantization functions (#5830)
2024-03-02 14:19:09 +00:00
size_t local_size = 0;
while (true) {
std::unique_lock<std::mutex> lock(mutex);
llama : add Command R Plus support (#6491) * Add Command R Plus GGUF * Add Command R Plus GGUF * Loading works up to LayerNorm2D * Export new tensors in 1D so they are not quantized. * Fix embedding layer based on Noeda's example * Whitespace * Add line * Fix unexpected tokens on MPS. Re-add F16 fix. ((Noeda) * dranger003: Fix block index overflow in CUDA dequantizing. * Reverted blocked multiplication code as it still has issues and could affect other Llama arches * export norms as f32 * fix overflow issues during quant and other cleanup * Type convention Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * dranger003: Fix more int overflow during quant. --------- Co-authored-by: S <seast@Ss-Mac-Studio.local> Co-authored-by: S <s@example.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-09 08:16:13 +00:00
int64_t first_row = counter; counter += nrows_per_chunk;
llama : refactor internal quantization functions (#5830)
2024-03-02 14:19:09 +00:00
if (first_row >= nrows) {
if (local_size > 0) {
new_size += local_size;
}
break;
}
lock.unlock();
llama : add Command R Plus support (#6491) * Add Command R Plus GGUF * Add Command R Plus GGUF * Loading works up to LayerNorm2D * Export new tensors in 1D so they are not quantized. * Fix embedding layer based on Noeda's example * Whitespace * Add line * Fix unexpected tokens on MPS. Re-add F16 fix. ((Noeda) * dranger003: Fix block index overflow in CUDA dequantizing. * Reverted blocked multiplication code as it still has issues and could affect other Llama arches * export norms as f32 * fix overflow issues during quant and other cleanup * Type convention Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * dranger003: Fix more int overflow during quant. --------- Co-authored-by: S <seast@Ss-Mac-Studio.local> Co-authored-by: S <s@example.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-09 08:16:13 +00:00
const int64_t this_nrow = std::min(nrows - first_row, nrows_per_chunk);
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
size_t this_size = ggml_quantize_chunk(new_type, f32_data, new_data, first_row * n_per_row, this_nrow, n_per_row, imatrix);
local_size += this_size;
// validate the quantized data
const size_t row_size = ggml_row_size(new_type, n_per_row);
void * this_data = (char *) new_data + first_row * row_size;
if (!ggml_validate_row_data(new_type, this_data, this_size)) {
std::unique_lock<std::mutex> lock(mutex);
valid = false;
break;
}
llama : refactor internal quantization functions (#5830)
2024-03-02 14:19:09 +00:00
}
};
for (int it = 0; it < nthread - 1; ++it) {
workers.emplace_back(compute);
}
compute();
for (auto & w : workers) { w.join(); }
workers.clear();
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
if (!valid) {
throw std::runtime_error("quantized data validation failed");
}
llama : refactor internal quantization functions (#5830)
2024-03-02 14:19:09 +00:00
return new_size;
}
llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691) * Add support for quantizing already quantized models * Threaded dequantizing and f16 to f32 conversion * Clean up thread blocks with spares calculation a bit * Use std::runtime_error exceptions.
2023-06-10 07:59:17 +00:00
static void llama_model_quantize_internal(const std::string & fname_inp, const std::string & fname_out, const llama_model_quantize_params * params) {
llama : fix F16/F32 downcast + improve names (#5980)
2024-03-11 07:56:47 +00:00
ggml_type default_type;
llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691) * Add support for quantizing already quantized models * Threaded dequantizing and f16 to f32 conversion * Clean up thread blocks with spares calculation a bit * Use std::runtime_error exceptions.
2023-06-10 07:59:17 +00:00
llama_ftype ftype = params->ftype;
switch (params->ftype) {
llama : fix F16/F32 downcast + improve names (#5980)
2024-03-11 07:56:47 +00:00
case LLAMA_FTYPE_MOSTLY_Q4_0: default_type = GGML_TYPE_Q4_0; break;
case LLAMA_FTYPE_MOSTLY_Q4_1: default_type = GGML_TYPE_Q4_1; break;
case LLAMA_FTYPE_MOSTLY_Q5_0: default_type = GGML_TYPE_Q5_0; break;
case LLAMA_FTYPE_MOSTLY_Q5_1: default_type = GGML_TYPE_Q5_1; break;
case LLAMA_FTYPE_MOSTLY_Q8_0: default_type = GGML_TYPE_Q8_0; break;
case LLAMA_FTYPE_MOSTLY_F16: default_type = GGML_TYPE_F16; break;
case LLAMA_FTYPE_ALL_F32: default_type = GGML_TYPE_F32; break;
llama : consistently catch and throw only exceptions deriving from std::exception (#1599) Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-05 20:24:29 +00:00
ggml : add SOTA 2,3,4,5,6 bit k-quantizations (#1684) * Starting to add k-quantization to ggml I think it is better to have quantization separate from ggml. For now just adding the k-quants there, but it would be better to also factor out the existing ggml quantizations. * Adding Q3_K and Q8_K (de)-quantization * Q3_K now working on CUDA and AVX2/scalar CUDA is not ideal - ~50% slower than Q4_0 for single token prediction, about the same in batch mode (perplexity). CPU single token is ~55 ms (on Ryzen 7950X). * Some improvement for Q3_K on CUDA It is now ~22.5 ms/token on my GPU, so ~30% slower than Q4_0. * Some more CUDA optimizations for Q3_K Single token is now 20.5 ms/token (~20% slower than Q4_0). Perplexity is on par with Q4_0. * Adding Q4_K - scalar, AVX2, CUDA Performance is the same or perhaps very slightly better than Q4_0 on the CPU. On the GPU, single token prediction is ~10% better than Q4_0, batch mode (perplexity is about the same). * Adding Q6_K - scalar, AVX2, CUDA Performance is ~40% lower compared to Q4_K on the CPU. This is to be expected, considering that we are memory bound on the CPU and the 6-bit model is ~44% larger than the 4-bit. On the GPU, single token prediction is ~6% lower than Q4_0, batch mode (perplexity) is even closer (but still slower). * Adding Q5_K - scalar, AVX2, CUDA Performance is ~20% lower compared to Q4_K on the CPU. This is to be expected, considering that we are memory bound on the CPU and the 5-bit model is ~22% larger than the 4-bit. On the GPU, single token prediction is about the same as Q4_0 for both, single token and batch prediction. * Per convention, all QX_K quantizations use Q5_K for output.weight * Adding quantization mixes * Quantization mixes: didn't quite get what I wanted in the last commit * Q4_K dot product for ARM_NEON * Q6_K dot product for ARM_NEON * Q5_K dot product for ARM_NEON * Adding Q3_K dot for ARM_NEON It is 22% slower than Q4_K, despite the smaller model size. On x86_64, where we are memory bound, the Q3_K model is quite a bit faster than Q4_K. * A very slightly faster ARM_NEON Q3_K dot * Adding Q2_K - just CUDA for now Token prediction is pretty good - about 15.5 ms on a RTX 4080. Perplexity is about the same as Q4_K. * Adding scalar and AVX2 Q2_K dot * Adding ARM_NEON Q2_K dot About the same performance as Q4_K. * A slightly faster ARM_NEON Q2_K dot Single token prediction is now ~36 ms on M2 Max. The code is much simpler too. * Fixed bug in Q2_K CUDA dot product kernel Stranegly enough, for the few prompts I tried with the 7B model the responses looked perfectly reasonable. Only realized something is not quite right when I tried the larger models and started getting nonse back. In any case, Q2_K single token evaluation time on an RTX 4080 in a Ryzen7950X box iusing CUDA and model fully loaded on the GPU are ~15.5 ms for 7B, ~25.4 ms for 13B, and ~55.8 ms for 30B. The max number of layers that fit in VRAM for The 65B is 32. With that, we get ~330 ms per token, which is not that much faster than just running on the CPU (~470 ms per token). * Don't print zeros/NaNs when no count histogram has been collected * A 10% faster CUDA vector dot kernel for Q3_K Q3_K is now running at ~18.5 ms / token on CUDA, so the gap to Q4_0 is only 10%. It seems memory acccess pattern is more important for performance than the amount of computation the kernel does. * A slightly daster Q4_K AVX2 dot product For perplexity, where we are less memory bound, time per pass drops by ~5%. Barely measurable difference for single token prediction. * A slightly faster ARM_NEON A4_K dot product * Minor * Fix quantization error test We cannot possibly be expecting rmse < 0.002 for 2- and 3-bit quantization variants. * Fix docker build I have been sloppy with vector reinterpret casts on ARM_NEON. It seems clang is very forgiving in that regard. * Added forgotten ggml.o dependence on k_quants.h to the Makefile * Had unintentionally committed the Makefile with -Ofast enabled * ggml : rename k_quants -> ggml-quants-k, use lowercase in code --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-05 19:56:18 +00:00
// K-quants
llama : add Q3_K_XS (#5060) * Add Q3_K_XS - intermediate size between Q2_K and Q3_K_S * Q3_K_XS: quanize first 1/8 of ffn_down layers with Q4_K Together with an importance matrix, this brings perplexity for LLaMA-v2-70B below the perplexity of the former Q2_K with a 800 MB smaller quantized model size. --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-22 10:43:33 +00:00
case LLAMA_FTYPE_MOSTLY_Q2_K_S:
llama : fix F16/F32 downcast + improve names (#5980)
2024-03-11 07:56:47 +00:00
case LLAMA_FTYPE_MOSTLY_Q2_K: default_type = GGML_TYPE_Q2_K; break;
case LLAMA_FTYPE_MOSTLY_IQ3_XS: default_type = GGML_TYPE_IQ3_S; break;
ggml : add SOTA 2,3,4,5,6 bit k-quantizations (#1684) * Starting to add k-quantization to ggml I think it is better to have quantization separate from ggml. For now just adding the k-quants there, but it would be better to also factor out the existing ggml quantizations. * Adding Q3_K and Q8_K (de)-quantization * Q3_K now working on CUDA and AVX2/scalar CUDA is not ideal - ~50% slower than Q4_0 for single token prediction, about the same in batch mode (perplexity). CPU single token is ~55 ms (on Ryzen 7950X). * Some improvement for Q3_K on CUDA It is now ~22.5 ms/token on my GPU, so ~30% slower than Q4_0. * Some more CUDA optimizations for Q3_K Single token is now 20.5 ms/token (~20% slower than Q4_0). Perplexity is on par with Q4_0. * Adding Q4_K - scalar, AVX2, CUDA Performance is the same or perhaps very slightly better than Q4_0 on the CPU. On the GPU, single token prediction is ~10% better than Q4_0, batch mode (perplexity is about the same). * Adding Q6_K - scalar, AVX2, CUDA Performance is ~40% lower compared to Q4_K on the CPU. This is to be expected, considering that we are memory bound on the CPU and the 6-bit model is ~44% larger than the 4-bit. On the GPU, single token prediction is ~6% lower than Q4_0, batch mode (perplexity) is even closer (but still slower). * Adding Q5_K - scalar, AVX2, CUDA Performance is ~20% lower compared to Q4_K on the CPU. This is to be expected, considering that we are memory bound on the CPU and the 5-bit model is ~22% larger than the 4-bit. On the GPU, single token prediction is about the same as Q4_0 for both, single token and batch prediction. * Per convention, all QX_K quantizations use Q5_K for output.weight * Adding quantization mixes * Quantization mixes: didn't quite get what I wanted in the last commit * Q4_K dot product for ARM_NEON * Q6_K dot product for ARM_NEON * Q5_K dot product for ARM_NEON * Adding Q3_K dot for ARM_NEON It is 22% slower than Q4_K, despite the smaller model size. On x86_64, where we are memory bound, the Q3_K model is quite a bit faster than Q4_K. * A very slightly faster ARM_NEON Q3_K dot * Adding Q2_K - just CUDA for now Token prediction is pretty good - about 15.5 ms on a RTX 4080. Perplexity is about the same as Q4_K. * Adding scalar and AVX2 Q2_K dot * Adding ARM_NEON Q2_K dot About the same performance as Q4_K. * A slightly faster ARM_NEON Q2_K dot Single token prediction is now ~36 ms on M2 Max. The code is much simpler too. * Fixed bug in Q2_K CUDA dot product kernel Stranegly enough, for the few prompts I tried with the 7B model the responses looked perfectly reasonable. Only realized something is not quite right when I tried the larger models and started getting nonse back. In any case, Q2_K single token evaluation time on an RTX 4080 in a Ryzen7950X box iusing CUDA and model fully loaded on the GPU are ~15.5 ms for 7B, ~25.4 ms for 13B, and ~55.8 ms for 30B. The max number of layers that fit in VRAM for The 65B is 32. With that, we get ~330 ms per token, which is not that much faster than just running on the CPU (~470 ms per token). * Don't print zeros/NaNs when no count histogram has been collected * A 10% faster CUDA vector dot kernel for Q3_K Q3_K is now running at ~18.5 ms / token on CUDA, so the gap to Q4_0 is only 10%. It seems memory acccess pattern is more important for performance than the amount of computation the kernel does. * A slightly daster Q4_K AVX2 dot product For perplexity, where we are less memory bound, time per pass drops by ~5%. Barely measurable difference for single token prediction. * A slightly faster ARM_NEON A4_K dot product * Minor * Fix quantization error test We cannot possibly be expecting rmse < 0.002 for 2- and 3-bit quantization variants. * Fix docker build I have been sloppy with vector reinterpret casts on ARM_NEON. It seems clang is very forgiving in that regard. * Added forgotten ggml.o dependence on k_quants.h to the Makefile * Had unintentionally committed the Makefile with -Ofast enabled * ggml : rename k_quants -> ggml-quants-k, use lowercase in code --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-05 19:56:18 +00:00
case LLAMA_FTYPE_MOSTLY_Q3_K_S:
case LLAMA_FTYPE_MOSTLY_Q3_K_M:
llama : fix F16/F32 downcast + improve names (#5980)
2024-03-11 07:56:47 +00:00
case LLAMA_FTYPE_MOSTLY_Q3_K_L: default_type = GGML_TYPE_Q3_K; break;
ggml : add SOTA 2,3,4,5,6 bit k-quantizations (#1684) * Starting to add k-quantization to ggml I think it is better to have quantization separate from ggml. For now just adding the k-quants there, but it would be better to also factor out the existing ggml quantizations. * Adding Q3_K and Q8_K (de)-quantization * Q3_K now working on CUDA and AVX2/scalar CUDA is not ideal - ~50% slower than Q4_0 for single token prediction, about the same in batch mode (perplexity). CPU single token is ~55 ms (on Ryzen 7950X). * Some improvement for Q3_K on CUDA It is now ~22.5 ms/token on my GPU, so ~30% slower than Q4_0. * Some more CUDA optimizations for Q3_K Single token is now 20.5 ms/token (~20% slower than Q4_0). Perplexity is on par with Q4_0. * Adding Q4_K - scalar, AVX2, CUDA Performance is the same or perhaps very slightly better than Q4_0 on the CPU. On the GPU, single token prediction is ~10% better than Q4_0, batch mode (perplexity is about the same). * Adding Q6_K - scalar, AVX2, CUDA Performance is ~40% lower compared to Q4_K on the CPU. This is to be expected, considering that we are memory bound on the CPU and the 6-bit model is ~44% larger than the 4-bit. On the GPU, single token prediction is ~6% lower than Q4_0, batch mode (perplexity) is even closer (but still slower). * Adding Q5_K - scalar, AVX2, CUDA Performance is ~20% lower compared to Q4_K on the CPU. This is to be expected, considering that we are memory bound on the CPU and the 5-bit model is ~22% larger than the 4-bit. On the GPU, single token prediction is about the same as Q4_0 for both, single token and batch prediction. * Per convention, all QX_K quantizations use Q5_K for output.weight * Adding quantization mixes * Quantization mixes: didn't quite get what I wanted in the last commit * Q4_K dot product for ARM_NEON * Q6_K dot product for ARM_NEON * Q5_K dot product for ARM_NEON * Adding Q3_K dot for ARM_NEON It is 22% slower than Q4_K, despite the smaller model size. On x86_64, where we are memory bound, the Q3_K model is quite a bit faster than Q4_K. * A very slightly faster ARM_NEON Q3_K dot * Adding Q2_K - just CUDA for now Token prediction is pretty good - about 15.5 ms on a RTX 4080. Perplexity is about the same as Q4_K. * Adding scalar and AVX2 Q2_K dot * Adding ARM_NEON Q2_K dot About the same performance as Q4_K. * A slightly faster ARM_NEON Q2_K dot Single token prediction is now ~36 ms on M2 Max. The code is much simpler too. * Fixed bug in Q2_K CUDA dot product kernel Stranegly enough, for the few prompts I tried with the 7B model the responses looked perfectly reasonable. Only realized something is not quite right when I tried the larger models and started getting nonse back. In any case, Q2_K single token evaluation time on an RTX 4080 in a Ryzen7950X box iusing CUDA and model fully loaded on the GPU are ~15.5 ms for 7B, ~25.4 ms for 13B, and ~55.8 ms for 30B. The max number of layers that fit in VRAM for The 65B is 32. With that, we get ~330 ms per token, which is not that much faster than just running on the CPU (~470 ms per token). * Don't print zeros/NaNs when no count histogram has been collected * A 10% faster CUDA vector dot kernel for Q3_K Q3_K is now running at ~18.5 ms / token on CUDA, so the gap to Q4_0 is only 10%. It seems memory acccess pattern is more important for performance than the amount of computation the kernel does. * A slightly daster Q4_K AVX2 dot product For perplexity, where we are less memory bound, time per pass drops by ~5%. Barely measurable difference for single token prediction. * A slightly faster ARM_NEON A4_K dot product * Minor * Fix quantization error test We cannot possibly be expecting rmse < 0.002 for 2- and 3-bit quantization variants. * Fix docker build I have been sloppy with vector reinterpret casts on ARM_NEON. It seems clang is very forgiving in that regard. * Added forgotten ggml.o dependence on k_quants.h to the Makefile * Had unintentionally committed the Makefile with -Ofast enabled * ggml : rename k_quants -> ggml-quants-k, use lowercase in code --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-05 19:56:18 +00:00
case LLAMA_FTYPE_MOSTLY_Q4_K_S:
llama : fix F16/F32 downcast + improve names (#5980)
2024-03-11 07:56:47 +00:00
case LLAMA_FTYPE_MOSTLY_Q4_K_M: default_type = GGML_TYPE_Q4_K; break;
ggml : add SOTA 2,3,4,5,6 bit k-quantizations (#1684) * Starting to add k-quantization to ggml I think it is better to have quantization separate from ggml. For now just adding the k-quants there, but it would be better to also factor out the existing ggml quantizations. * Adding Q3_K and Q8_K (de)-quantization * Q3_K now working on CUDA and AVX2/scalar CUDA is not ideal - ~50% slower than Q4_0 for single token prediction, about the same in batch mode (perplexity). CPU single token is ~55 ms (on Ryzen 7950X). * Some improvement for Q3_K on CUDA It is now ~22.5 ms/token on my GPU, so ~30% slower than Q4_0. * Some more CUDA optimizations for Q3_K Single token is now 20.5 ms/token (~20% slower than Q4_0). Perplexity is on par with Q4_0. * Adding Q4_K - scalar, AVX2, CUDA Performance is the same or perhaps very slightly better than Q4_0 on the CPU. On the GPU, single token prediction is ~10% better than Q4_0, batch mode (perplexity is about the same). * Adding Q6_K - scalar, AVX2, CUDA Performance is ~40% lower compared to Q4_K on the CPU. This is to be expected, considering that we are memory bound on the CPU and the 6-bit model is ~44% larger than the 4-bit. On the GPU, single token prediction is ~6% lower than Q4_0, batch mode (perplexity) is even closer (but still slower). * Adding Q5_K - scalar, AVX2, CUDA Performance is ~20% lower compared to Q4_K on the CPU. This is to be expected, considering that we are memory bound on the CPU and the 5-bit model is ~22% larger than the 4-bit. On the GPU, single token prediction is about the same as Q4_0 for both, single token and batch prediction. * Per convention, all QX_K quantizations use Q5_K for output.weight * Adding quantization mixes * Quantization mixes: didn't quite get what I wanted in the last commit * Q4_K dot product for ARM_NEON * Q6_K dot product for ARM_NEON * Q5_K dot product for ARM_NEON * Adding Q3_K dot for ARM_NEON It is 22% slower than Q4_K, despite the smaller model size. On x86_64, where we are memory bound, the Q3_K model is quite a bit faster than Q4_K. * A very slightly faster ARM_NEON Q3_K dot * Adding Q2_K - just CUDA for now Token prediction is pretty good - about 15.5 ms on a RTX 4080. Perplexity is about the same as Q4_K. * Adding scalar and AVX2 Q2_K dot * Adding ARM_NEON Q2_K dot About the same performance as Q4_K. * A slightly faster ARM_NEON Q2_K dot Single token prediction is now ~36 ms on M2 Max. The code is much simpler too. * Fixed bug in Q2_K CUDA dot product kernel Stranegly enough, for the few prompts I tried with the 7B model the responses looked perfectly reasonable. Only realized something is not quite right when I tried the larger models and started getting nonse back. In any case, Q2_K single token evaluation time on an RTX 4080 in a Ryzen7950X box iusing CUDA and model fully loaded on the GPU are ~15.5 ms for 7B, ~25.4 ms for 13B, and ~55.8 ms for 30B. The max number of layers that fit in VRAM for The 65B is 32. With that, we get ~330 ms per token, which is not that much faster than just running on the CPU (~470 ms per token). * Don't print zeros/NaNs when no count histogram has been collected * A 10% faster CUDA vector dot kernel for Q3_K Q3_K is now running at ~18.5 ms / token on CUDA, so the gap to Q4_0 is only 10%. It seems memory acccess pattern is more important for performance than the amount of computation the kernel does. * A slightly daster Q4_K AVX2 dot product For perplexity, where we are less memory bound, time per pass drops by ~5%. Barely measurable difference for single token prediction. * A slightly faster ARM_NEON A4_K dot product * Minor * Fix quantization error test We cannot possibly be expecting rmse < 0.002 for 2- and 3-bit quantization variants. * Fix docker build I have been sloppy with vector reinterpret casts on ARM_NEON. It seems clang is very forgiving in that regard. * Added forgotten ggml.o dependence on k_quants.h to the Makefile * Had unintentionally committed the Makefile with -Ofast enabled * ggml : rename k_quants -> ggml-quants-k, use lowercase in code --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-05 19:56:18 +00:00
case LLAMA_FTYPE_MOSTLY_Q5_K_S:
llama : fix F16/F32 downcast + improve names (#5980)
2024-03-11 07:56:47 +00:00
case LLAMA_FTYPE_MOSTLY_Q5_K_M: default_type = GGML_TYPE_Q5_K; break;
case LLAMA_FTYPE_MOSTLY_Q6_K: default_type = GGML_TYPE_Q6_K; break;
case LLAMA_FTYPE_MOSTLY_IQ2_XXS: default_type = GGML_TYPE_IQ2_XXS; break;
case LLAMA_FTYPE_MOSTLY_IQ2_XS: default_type = GGML_TYPE_IQ2_XS; break;
case LLAMA_FTYPE_MOSTLY_IQ2_S: default_type = GGML_TYPE_IQ2_XS; break;
case LLAMA_FTYPE_MOSTLY_IQ2_M: default_type = GGML_TYPE_IQ2_S; break;
case LLAMA_FTYPE_MOSTLY_IQ3_XXS: default_type = GGML_TYPE_IQ3_XXS; break;
case LLAMA_FTYPE_MOSTLY_IQ1_S: default_type = GGML_TYPE_IQ1_S; break;
IQ1_M: 1.75 bpw quantization (#6302) * iq1_m: basics * iq1_m: basics-2 * iq1_m: CUDA dequantize works Very 1st shot I get PPL = 9.76 for LLaMA-v2-7B. * iq1_m: separate shifts for each group of 8 in a block We get PPL(LLaMA-v2-7B ) = 9.2810 PPL(LLaMA-v2-13B) = 6.8105 Not bad, but slightly higher than sqrt(PPL(IQ1_S) * PPL(IQ2_XXS)) which is the expected outcome given that IQ1_M is halfway between IQ1_S and IQ2_XXS in terms of bpw. From this, we would expect PPL = 9.14 for LLaMA-v2-7B PPL = 6.63 for LLaMA-v2-13B * iq1_m: go to 3-bit scales There is slight increase in PPL, but the 0.0625 bpw reduction in size is totally worth it. We now have PPL(LLaMA-v2-7B ) = 9.4469 at 1.96 bpw PPL(LLaMA-v2-13B) = 6.8717 at 1.93 bpw PPL(LLaMA-v2-70B) = 4.8568 at 1.85 bpw * iq1_m: scalar dot product * iq1_m: AVX2 dot product * iq1_m: very slightly faster AVX2 dot product * iq1_m: ARM_NEON dot product Works, but very slow (10.5 t/s) * iq1_m: Metal - dequantize works, dot product does not * iq1_m: Metal now works About the same performance as iq1_s. * iq1_m: minor * iq1_m: checking pure iq1_m quantization It is pretty bad: PPL(LLaMA-v2-7B) = 34 if we quantize output.weight with Q4_K. * iiq1_m: slightly faster ARM_NEON dot product 10.5 t/s -> 11.65 t/s * iq1_m: faster ARM_NEON dot product 11.65 t/s -> 14.9 t/s * iq1_m: another minor ARM_NEON dot product improvement 14.9 -> 15.0 t/s * iq1_m: small PPL improvement via super-block scale adjustment After quantizing block scales redo the super-block scale fit. PPL(LLaMA-v2-7B ) = 9.3346 PPL(LLaMA-v2-13B) = 6.8419 PPL(LLaMA-v2-70B) = 4.8294 PPL(Mistral-7B ) = 8.1624 * iq1_m: adapt to CUDA refactoring * iq1_m: remove unused variable We have progressed to warnings being errors. * iq1_m: add to backend-ops tests * iq1_m: fix Windows ARM * iq1_m: use common definition of iq1m_scale_t * cuda: assert -> NO_DEVICE_CODE * iq1_M: PR comments --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-26 14:21:27 +00:00
case LLAMA_FTYPE_MOSTLY_IQ1_M: default_type = GGML_TYPE_IQ1_M; break;
llama : fix F16/F32 downcast + improve names (#5980)
2024-03-11 07:56:47 +00:00
case LLAMA_FTYPE_MOSTLY_IQ4_NL: default_type = GGML_TYPE_IQ4_NL; break;
case LLAMA_FTYPE_MOSTLY_IQ4_XS: default_type = GGML_TYPE_IQ4_XS; break;
case LLAMA_FTYPE_MOSTLY_IQ3_S: default_type = GGML_TYPE_IQ3_S; break;
case LLAMA_FTYPE_MOSTLY_IQ3_M: default_type = GGML_TYPE_IQ3_S; break;
ggml : quantization refactoring (#3833) * ggml : factor all quantization code in ggml-quants ggml-ci * ggml-quants : fix Zig and Swift builds + quantize tool ggml-ci * quantize : --pure option for disabling k-quant mixtures --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com>
2023-10-29 16:32:28 +00:00
llama : consistently catch and throw only exceptions deriving from std::exception (#1599) Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-05 20:24:29 +00:00
default: throw std::runtime_error(format("invalid output file type %d\n", ftype));
ggml : add SOTA 2,3,4,5,6 bit k-quantizations (#1684) * Starting to add k-quantization to ggml I think it is better to have quantization separate from ggml. For now just adding the k-quants there, but it would be better to also factor out the existing ggml quantizations. * Adding Q3_K and Q8_K (de)-quantization * Q3_K now working on CUDA and AVX2/scalar CUDA is not ideal - ~50% slower than Q4_0 for single token prediction, about the same in batch mode (perplexity). CPU single token is ~55 ms (on Ryzen 7950X). * Some improvement for Q3_K on CUDA It is now ~22.5 ms/token on my GPU, so ~30% slower than Q4_0. * Some more CUDA optimizations for Q3_K Single token is now 20.5 ms/token (~20% slower than Q4_0). Perplexity is on par with Q4_0. * Adding Q4_K - scalar, AVX2, CUDA Performance is the same or perhaps very slightly better than Q4_0 on the CPU. On the GPU, single token prediction is ~10% better than Q4_0, batch mode (perplexity is about the same). * Adding Q6_K - scalar, AVX2, CUDA Performance is ~40% lower compared to Q4_K on the CPU. This is to be expected, considering that we are memory bound on the CPU and the 6-bit model is ~44% larger than the 4-bit. On the GPU, single token prediction is ~6% lower than Q4_0, batch mode (perplexity) is even closer (but still slower). * Adding Q5_K - scalar, AVX2, CUDA Performance is ~20% lower compared to Q4_K on the CPU. This is to be expected, considering that we are memory bound on the CPU and the 5-bit model is ~22% larger than the 4-bit. On the GPU, single token prediction is about the same as Q4_0 for both, single token and batch prediction. * Per convention, all QX_K quantizations use Q5_K for output.weight * Adding quantization mixes * Quantization mixes: didn't quite get what I wanted in the last commit * Q4_K dot product for ARM_NEON * Q6_K dot product for ARM_NEON * Q5_K dot product for ARM_NEON * Adding Q3_K dot for ARM_NEON It is 22% slower than Q4_K, despite the smaller model size. On x86_64, where we are memory bound, the Q3_K model is quite a bit faster than Q4_K. * A very slightly faster ARM_NEON Q3_K dot * Adding Q2_K - just CUDA for now Token prediction is pretty good - about 15.5 ms on a RTX 4080. Perplexity is about the same as Q4_K. * Adding scalar and AVX2 Q2_K dot * Adding ARM_NEON Q2_K dot About the same performance as Q4_K. * A slightly faster ARM_NEON Q2_K dot Single token prediction is now ~36 ms on M2 Max. The code is much simpler too. * Fixed bug in Q2_K CUDA dot product kernel Stranegly enough, for the few prompts I tried with the 7B model the responses looked perfectly reasonable. Only realized something is not quite right when I tried the larger models and started getting nonse back. In any case, Q2_K single token evaluation time on an RTX 4080 in a Ryzen7950X box iusing CUDA and model fully loaded on the GPU are ~15.5 ms for 7B, ~25.4 ms for 13B, and ~55.8 ms for 30B. The max number of layers that fit in VRAM for The 65B is 32. With that, we get ~330 ms per token, which is not that much faster than just running on the CPU (~470 ms per token). * Don't print zeros/NaNs when no count histogram has been collected * A 10% faster CUDA vector dot kernel for Q3_K Q3_K is now running at ~18.5 ms / token on CUDA, so the gap to Q4_0 is only 10%. It seems memory acccess pattern is more important for performance than the amount of computation the kernel does. * A slightly daster Q4_K AVX2 dot product For perplexity, where we are less memory bound, time per pass drops by ~5%. Barely measurable difference for single token prediction. * A slightly faster ARM_NEON A4_K dot product * Minor * Fix quantization error test We cannot possibly be expecting rmse < 0.002 for 2- and 3-bit quantization variants. * Fix docker build I have been sloppy with vector reinterpret casts on ARM_NEON. It seems clang is very forgiving in that regard. * Added forgotten ggml.o dependence on k_quants.h to the Makefile * Had unintentionally committed the Makefile with -Ofast enabled * ggml : rename k_quants -> ggml-quants-k, use lowercase in code --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-05 19:56:18 +00:00
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
int nthread = params->nthread;
llama : multi-threaded quantization (#1075) * Multi-threading quantization. Not much gain for simple quantizations, bit it will be important for quantizations that require more CPU cycles. * Multi-threading for quantize-stats It now does the job in ~14 seconds on my Mac for Q4_0, Q4_1 and Q4_2. Single-threaded it was taking more than 2 minutes after adding the more elaborate version of Q4_2. * Reviewer comments * Avoiding compiler confusion After changing chunk_size to const int as suggested by @ggerganov, clang and GCC starting to warn me that I don't need to capture it in the lambda. So, I removed it from the capture list. But that makes the MSVC build fail. So, making it a constexpr to make every compiler happy. * Still fighting with lambda captures in MSVC --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-04-20 17:42:27 +00:00
if (nthread <= 0) {
nthread = std::thread::hardware_concurrency();
}
llama : quantize up to 31% faster on Linux and Windows with mmap (#3206) * llama : enable mmap in quantize on Linux -> 31% faster * also enable mmap on Windows --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-29 13:48:45 +00:00
// mmap consistently increases speed Linux, and also increases speed on Windows with
// hot cache. It may cause a slowdown on macOS, possibly related to free memory.
#if defined(__linux__) || defined(_WIN32)
constexpr bool use_mmap = true;
#else
constexpr bool use_mmap = false;
#endif
quantize : be able to override metadata by key (#6321) * quantize: be able to override metadata by key * minor : spacing --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 12:09:30 +00:00
llama_model_kv_override * kv_overrides = nullptr;
if (params->kv_overrides) {
auto v = (std::vector<llama_model_kv_override>*)params->kv_overrides;
kv_overrides = v->data();
}
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
llama_model_loader ml(fname_inp, use_mmap, /*check_tensors*/ true, kv_overrides);
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
ml.init_mappings(false); // no prefetching
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
Better perplexity for 2- and 3-bit quantization for LLaMA-v2-70B (#2807) * Better perplexity for 2- and 3-bit quantization for the 70B model * PR comment --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2023-08-26 14:27:49 +00:00
llama_model model;
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
llm_load_arch(ml, model);
llm_load_hparams(ml, model);
Better perplexity for 2- and 3-bit quantization for LLaMA-v2-70B (#2807) * Better perplexity for 2- and 3-bit quantization for the 70B model * PR comment --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2023-08-26 14:27:49 +00:00
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
struct quantize_state_internal qs(model, params);
Allow quantize to only copy tensors, some other improvements (#2931) * Allow quantize tool to only copy tensors to allow repackaging models. * Slightly better logic when requantizing. * Change help message to go to `stdout`.
2023-09-01 14:02:48 +00:00
if (params->only_copy) {
ftype = model.ftype;
}
2-bit quantizations (#4897) * imatrix: load * imatrix: WIP * imatrix: Add Q2_K quantization * imatrix: also guard against Q2_K_S quantization without importance matrix * imatrix: guard even more against low-bit quantization misuse --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:45:56 +00:00
const std::unordered_map<std::string, std::vector<float>> * imatrix_data = nullptr;
if (params->imatrix) {
imatrix_data = static_cast<const std::unordered_map<std::string, std::vector<float>>*>(params->imatrix);
if (imatrix_data) {
llama : use LLAMA_LOG_ macros for logging
2024-01-14 09:03:19 +00:00
LLAMA_LOG_INFO("================================ Have weights data with %d entries\n",int(imatrix_data->size()));
ggml : importance matrix support for legacy quants (#4969) * imatrix: adding support for legacy quants * imatrix: guard Q4_0/Q5_0 against ffn_down craziness --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-16 17:51:26 +00:00
qs.has_imatrix = true;
2-bit quantizations (#4897) * imatrix: load * imatrix: WIP * imatrix: Add Q2_K quantization * imatrix: also guard against Q2_K_S quantization without importance matrix * imatrix: guard even more against low-bit quantization misuse --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:45:56 +00:00
}
}
Allow quantize to only copy tensors, some other improvements (#2931) * Allow quantize tool to only copy tensors to allow repackaging models. * Slightly better logic when requantizing. * Change help message to go to `stdout`.
2023-09-01 14:02:48 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
const size_t align = GGUF_DEFAULT_ALIGNMENT;
struct gguf_context * ctx_out = gguf_init_empty();
// copy the KV pairs from the input file
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
gguf_set_kv (ctx_out, ml.meta);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
gguf_set_val_u32(ctx_out, "general.quantization_version", GGML_QNT_VERSION);
gguf : add ftype meta info to the model (#2710) * llama : add ftype meta info to the model ggml-ci * convert.py : add ftype when converting (does not work) * convert.py : fix Enum to IntEnum ggml-ci
2023-08-22 17:05:59 +00:00
gguf_set_val_u32(ctx_out, "general.file_type", ftype);
llama : add gguf_remove_key + remove split meta during quantize (#6591) * Remove split metadata when quantize model shards * Find metadata key by enum * Correct loop range for gguf_remove_key and code format * Free kv memory --------- Co-authored-by: z5269887 <z5269887@unsw.edu.au>
2024-04-12 10:45:06 +00:00
// Remove split metadata
gguf_remove_key(ctx_out, ml.llm_kv(LLM_KV_SPLIT_NO).c_str());
gguf_remove_key(ctx_out, ml.llm_kv(LLM_KV_SPLIT_COUNT).c_str());
gguf_remove_key(ctx_out, ml.llm_kv(LLM_KV_SPLIT_TENSORS_COUNT).c_str());
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
quantize : be able to override metadata by key (#6321) * quantize: be able to override metadata by key * minor : spacing --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 12:09:30 +00:00
if (params->kv_overrides) {
const std::vector<llama_model_kv_override> & overrides = *(const std::vector<llama_model_kv_override> *)params->kv_overrides;
for (auto & o : overrides) {
if (o.key[0] == 0) break;
if (o.tag == LLAMA_KV_OVERRIDE_TYPE_FLOAT) {
quantize: add imatrix and dataset metadata in GGUF (#6658) * imatrix: save the dataset file used in the output file * llama: support kv overrides type string string * common: factorize KV Overrides parsing between common and server * quantize: add imatrix n entries and dataset KV metadata quantize: factorize KV Overrides parsing between common #6656 * llama: remove kv override str_value initialization as it does not compile on some toolchain * quantize: add imatrix m_last_call as `quantize.imatrix.chunks_count` * quantize: add imatrix filename in KV * llama: add llama_model_kv_override_free * common: add llama_model_kv_override_free common: free kv override if used after model loading * llama: finally move the string KV override value to the stack * llama : minor * no need to add a NUL to the std::vector, std::string can be initialized from a pair of iterators. Co-authored-by: slaren <slarengh@gmail.com> * kv override: ensure string termination --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-04-26 18:06:33 +00:00
gguf_set_val_f32(ctx_out, o.key, o.val_f64);
quantize : be able to override metadata by key (#6321) * quantize: be able to override metadata by key * minor : spacing --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 12:09:30 +00:00
} else if (o.tag == LLAMA_KV_OVERRIDE_TYPE_INT) {
quantize: add imatrix and dataset metadata in GGUF (#6658) * imatrix: save the dataset file used in the output file * llama: support kv overrides type string string * common: factorize KV Overrides parsing between common and server * quantize: add imatrix n entries and dataset KV metadata quantize: factorize KV Overrides parsing between common #6656 * llama: remove kv override str_value initialization as it does not compile on some toolchain * quantize: add imatrix m_last_call as `quantize.imatrix.chunks_count` * quantize: add imatrix filename in KV * llama: add llama_model_kv_override_free * common: add llama_model_kv_override_free common: free kv override if used after model loading * llama: finally move the string KV override value to the stack * llama : minor * no need to add a NUL to the std::vector, std::string can be initialized from a pair of iterators. Co-authored-by: slaren <slarengh@gmail.com> * kv override: ensure string termination --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-04-26 18:06:33 +00:00
gguf_set_val_i32(ctx_out, o.key, o.val_i64);
quantize : be able to override metadata by key (#6321) * quantize: be able to override metadata by key * minor : spacing --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 12:09:30 +00:00
} else if (o.tag == LLAMA_KV_OVERRIDE_TYPE_BOOL) {
quantize: add imatrix and dataset metadata in GGUF (#6658) * imatrix: save the dataset file used in the output file * llama: support kv overrides type string string * common: factorize KV Overrides parsing between common and server * quantize: add imatrix n entries and dataset KV metadata quantize: factorize KV Overrides parsing between common #6656 * llama: remove kv override str_value initialization as it does not compile on some toolchain * quantize: add imatrix m_last_call as `quantize.imatrix.chunks_count` * quantize: add imatrix filename in KV * llama: add llama_model_kv_override_free * common: add llama_model_kv_override_free common: free kv override if used after model loading * llama: finally move the string KV override value to the stack * llama : minor * no need to add a NUL to the std::vector, std::string can be initialized from a pair of iterators. Co-authored-by: slaren <slarengh@gmail.com> * kv override: ensure string termination --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-04-26 18:06:33 +00:00
gguf_set_val_bool(ctx_out, o.key, o.val_bool);
} else if (o.tag == LLAMA_KV_OVERRIDE_TYPE_STR) {
gguf_set_val_str(ctx_out, o.key, o.val_str);
quantize : be able to override metadata by key (#6321) * quantize: be able to override metadata by key * minor : spacing --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 12:09:30 +00:00
} else {
LLAMA_LOG_WARN("%s: unknown KV override type for key %s\n", __func__, o.key);
}
}
}
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
for (int i = 0; i < ml.n_tensors; ++i) {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
const struct ggml_tensor * meta = ml.get_tensor_meta(i);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
const std::string name = ggml_get_name(meta);
// TODO: avoid hardcoded tensor names - use the TN_* constants
llama : fix attention layer count sanity check (#6550) * llama : fix attention layer count sanity check * llama : fix parentheses in attention layer count sanity check There was otherwise a warning when compiling. --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net>
2024-04-08 19:25:49 +00:00
if (name.find("attn_v.weight") != std::string::npos ||
name.find("attn_qkv.weight") != std::string::npos) {
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
++qs.n_attention_wv;
quantize : be able to override metadata by key (#6321) * quantize: be able to override metadata by key * minor : spacing --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 12:09:30 +00:00
} else if (name == LLM_TN(model.arch)(LLM_TENSOR_OUTPUT, "weight")) {
llama : fix quantization of shared token_embd (#5944)
2024-03-08 15:53:37 +00:00
qs.has_output = true;
}
ggml : add SOTA 2,3,4,5,6 bit k-quantizations (#1684) * Starting to add k-quantization to ggml I think it is better to have quantization separate from ggml. For now just adding the k-quants there, but it would be better to also factor out the existing ggml quantizations. * Adding Q3_K and Q8_K (de)-quantization * Q3_K now working on CUDA and AVX2/scalar CUDA is not ideal - ~50% slower than Q4_0 for single token prediction, about the same in batch mode (perplexity). CPU single token is ~55 ms (on Ryzen 7950X). * Some improvement for Q3_K on CUDA It is now ~22.5 ms/token on my GPU, so ~30% slower than Q4_0. * Some more CUDA optimizations for Q3_K Single token is now 20.5 ms/token (~20% slower than Q4_0). Perplexity is on par with Q4_0. * Adding Q4_K - scalar, AVX2, CUDA Performance is the same or perhaps very slightly better than Q4_0 on the CPU. On the GPU, single token prediction is ~10% better than Q4_0, batch mode (perplexity is about the same). * Adding Q6_K - scalar, AVX2, CUDA Performance is ~40% lower compared to Q4_K on the CPU. This is to be expected, considering that we are memory bound on the CPU and the 6-bit model is ~44% larger than the 4-bit. On the GPU, single token prediction is ~6% lower than Q4_0, batch mode (perplexity) is even closer (but still slower). * Adding Q5_K - scalar, AVX2, CUDA Performance is ~20% lower compared to Q4_K on the CPU. This is to be expected, considering that we are memory bound on the CPU and the 5-bit model is ~22% larger than the 4-bit. On the GPU, single token prediction is about the same as Q4_0 for both, single token and batch prediction. * Per convention, all QX_K quantizations use Q5_K for output.weight * Adding quantization mixes * Quantization mixes: didn't quite get what I wanted in the last commit * Q4_K dot product for ARM_NEON * Q6_K dot product for ARM_NEON * Q5_K dot product for ARM_NEON * Adding Q3_K dot for ARM_NEON It is 22% slower than Q4_K, despite the smaller model size. On x86_64, where we are memory bound, the Q3_K model is quite a bit faster than Q4_K. * A very slightly faster ARM_NEON Q3_K dot * Adding Q2_K - just CUDA for now Token prediction is pretty good - about 15.5 ms on a RTX 4080. Perplexity is about the same as Q4_K. * Adding scalar and AVX2 Q2_K dot * Adding ARM_NEON Q2_K dot About the same performance as Q4_K. * A slightly faster ARM_NEON Q2_K dot Single token prediction is now ~36 ms on M2 Max. The code is much simpler too. * Fixed bug in Q2_K CUDA dot product kernel Stranegly enough, for the few prompts I tried with the 7B model the responses looked perfectly reasonable. Only realized something is not quite right when I tried the larger models and started getting nonse back. In any case, Q2_K single token evaluation time on an RTX 4080 in a Ryzen7950X box iusing CUDA and model fully loaded on the GPU are ~15.5 ms for 7B, ~25.4 ms for 13B, and ~55.8 ms for 30B. The max number of layers that fit in VRAM for The 65B is 32. With that, we get ~330 ms per token, which is not that much faster than just running on the CPU (~470 ms per token). * Don't print zeros/NaNs when no count histogram has been collected * A 10% faster CUDA vector dot kernel for Q3_K Q3_K is now running at ~18.5 ms / token on CUDA, so the gap to Q4_0 is only 10%. It seems memory acccess pattern is more important for performance than the amount of computation the kernel does. * A slightly daster Q4_K AVX2 dot product For perplexity, where we are less memory bound, time per pass drops by ~5%. Barely measurable difference for single token prediction. * A slightly faster ARM_NEON A4_K dot product * Minor * Fix quantization error test We cannot possibly be expecting rmse < 0.002 for 2- and 3-bit quantization variants. * Fix docker build I have been sloppy with vector reinterpret casts on ARM_NEON. It seems clang is very forgiving in that regard. * Added forgotten ggml.o dependence on k_quants.h to the Makefile * Had unintentionally committed the Makefile with -Ofast enabled * ggml : rename k_quants -> ggml-quants-k, use lowercase in code --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-05 19:56:18 +00:00
}
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
qs.n_ffn_down = qs.n_ffn_gate = qs.n_ffn_up = (int)model.hparams.n_layer;
// sanity checks
llama : fix attention layer count sanity check (#6550) * llama : fix attention layer count sanity check * llama : fix parentheses in attention layer count sanity check There was otherwise a warning when compiling. --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net>
2024-04-08 19:25:49 +00:00
//
// - qs.n_attention_wv == 0 for Mamba models
// - qs.n_attention_wv == model.hparams.n_layer for Transformer models
//
GGML_ASSERT((qs.n_attention_wv == 0 || qs.n_attention_wv == (int)model.hparams.n_layer) && "n_attention_wv is unexpected");
ggml : add SOTA 2,3,4,5,6 bit k-quantizations (#1684) * Starting to add k-quantization to ggml I think it is better to have quantization separate from ggml. For now just adding the k-quants there, but it would be better to also factor out the existing ggml quantizations. * Adding Q3_K and Q8_K (de)-quantization * Q3_K now working on CUDA and AVX2/scalar CUDA is not ideal - ~50% slower than Q4_0 for single token prediction, about the same in batch mode (perplexity). CPU single token is ~55 ms (on Ryzen 7950X). * Some improvement for Q3_K on CUDA It is now ~22.5 ms/token on my GPU, so ~30% slower than Q4_0. * Some more CUDA optimizations for Q3_K Single token is now 20.5 ms/token (~20% slower than Q4_0). Perplexity is on par with Q4_0. * Adding Q4_K - scalar, AVX2, CUDA Performance is the same or perhaps very slightly better than Q4_0 on the CPU. On the GPU, single token prediction is ~10% better than Q4_0, batch mode (perplexity is about the same). * Adding Q6_K - scalar, AVX2, CUDA Performance is ~40% lower compared to Q4_K on the CPU. This is to be expected, considering that we are memory bound on the CPU and the 6-bit model is ~44% larger than the 4-bit. On the GPU, single token prediction is ~6% lower than Q4_0, batch mode (perplexity) is even closer (but still slower). * Adding Q5_K - scalar, AVX2, CUDA Performance is ~20% lower compared to Q4_K on the CPU. This is to be expected, considering that we are memory bound on the CPU and the 5-bit model is ~22% larger than the 4-bit. On the GPU, single token prediction is about the same as Q4_0 for both, single token and batch prediction. * Per convention, all QX_K quantizations use Q5_K for output.weight * Adding quantization mixes * Quantization mixes: didn't quite get what I wanted in the last commit * Q4_K dot product for ARM_NEON * Q6_K dot product for ARM_NEON * Q5_K dot product for ARM_NEON * Adding Q3_K dot for ARM_NEON It is 22% slower than Q4_K, despite the smaller model size. On x86_64, where we are memory bound, the Q3_K model is quite a bit faster than Q4_K. * A very slightly faster ARM_NEON Q3_K dot * Adding Q2_K - just CUDA for now Token prediction is pretty good - about 15.5 ms on a RTX 4080. Perplexity is about the same as Q4_K. * Adding scalar and AVX2 Q2_K dot * Adding ARM_NEON Q2_K dot About the same performance as Q4_K. * A slightly faster ARM_NEON Q2_K dot Single token prediction is now ~36 ms on M2 Max. The code is much simpler too. * Fixed bug in Q2_K CUDA dot product kernel Stranegly enough, for the few prompts I tried with the 7B model the responses looked perfectly reasonable. Only realized something is not quite right when I tried the larger models and started getting nonse back. In any case, Q2_K single token evaluation time on an RTX 4080 in a Ryzen7950X box iusing CUDA and model fully loaded on the GPU are ~15.5 ms for 7B, ~25.4 ms for 13B, and ~55.8 ms for 30B. The max number of layers that fit in VRAM for The 65B is 32. With that, we get ~330 ms per token, which is not that much faster than just running on the CPU (~470 ms per token). * Don't print zeros/NaNs when no count histogram has been collected * A 10% faster CUDA vector dot kernel for Q3_K Q3_K is now running at ~18.5 ms / token on CUDA, so the gap to Q4_0 is only 10%. It seems memory acccess pattern is more important for performance than the amount of computation the kernel does. * A slightly daster Q4_K AVX2 dot product For perplexity, where we are less memory bound, time per pass drops by ~5%. Barely measurable difference for single token prediction. * A slightly faster ARM_NEON A4_K dot product * Minor * Fix quantization error test We cannot possibly be expecting rmse < 0.002 for 2- and 3-bit quantization variants. * Fix docker build I have been sloppy with vector reinterpret casts on ARM_NEON. It seems clang is very forgiving in that regard. * Added forgotten ggml.o dependence on k_quants.h to the Makefile * Had unintentionally committed the Makefile with -Ofast enabled * ggml : rename k_quants -> ggml-quants-k, use lowercase in code --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-05 19:56:18 +00:00
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
size_t total_size_org = 0;
size_t total_size_new = 0;
llama : multi-threaded quantization (#1075) * Multi-threading quantization. Not much gain for simple quantizations, bit it will be important for quantizations that require more CPU cycles. * Multi-threading for quantize-stats It now does the job in ~14 seconds on my Mac for Q4_0, Q4_1 and Q4_2. Single-threaded it was taking more than 2 minutes after adding the more elaborate version of Q4_2. * Reviewer comments * Avoiding compiler confusion After changing chunk_size to const int as suggested by @ggerganov, clang and GCC starting to warn me that I don't need to capture it in the lambda. So, I removed it from the capture list. But that makes the MSVC build fail. So, making it a constexpr to make every compiler happy. * Still fighting with lambda captures in MSVC --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-04-20 17:42:27 +00:00
std::vector<std::thread> workers;
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
workers.reserve(nthread);
llama : multi-threaded quantization (#1075) * Multi-threading quantization. Not much gain for simple quantizations, bit it will be important for quantizations that require more CPU cycles. * Multi-threading for quantize-stats It now does the job in ~14 seconds on my Mac for Q4_0, Q4_1 and Q4_2. Single-threaded it was taking more than 2 minutes after adding the more elaborate version of Q4_2. * Reviewer comments * Avoiding compiler confusion After changing chunk_size to const int as suggested by @ggerganov, clang and GCC starting to warn me that I don't need to capture it in the lambda. So, I removed it from the capture list. But that makes the MSVC build fail. So, making it a constexpr to make every compiler happy. * Still fighting with lambda captures in MSVC --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-04-20 17:42:27 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
int idx = 0;
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
std::vector<no_init<uint8_t>> read_data;
std::vector<no_init<uint8_t>> work;
std::vector<no_init<float>> f32_conv_buf;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
quantize : add '--keep-split' to quantize model into shards (#6688) * Implement '--keep-split' to quantize model into several shards * Add test script * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Split model correctly even if tensor id is out-of-order * Update llama_model_quantize_params * Fix preci failures --------- Co-authored-by: z5269887 <z5269887@unsw.edu.au> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-25 10:29:35 +00:00
uint16_t n_split = 1;
// Assume split index is continuous
if (params->keep_split) {
for (int i = 0; i < ml.n_tensors; ++i) {
n_split = std::max(uint16_t(ml.get_weight(i)->idx+1), n_split);
}
}
std::vector<gguf_context*> ctx_outs(n_split, NULL);
ctx_outs[0] = ctx_out;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
// populate the original tensors so we get an initial meta data
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
for (int i = 0; i < ml.n_tensors; ++i) {
quantize : add '--keep-split' to quantize model into shards (#6688) * Implement '--keep-split' to quantize model into several shards * Add test script * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Split model correctly even if tensor id is out-of-order * Update llama_model_quantize_params * Fix preci failures --------- Co-authored-by: z5269887 <z5269887@unsw.edu.au> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-25 10:29:35 +00:00
auto weight = ml.get_weight(i);
uint16_t i_split = params->keep_split ? weight->idx : 0;
struct ggml_tensor * tensor = weight->tensor;
if (ctx_outs[i_split] == NULL) {
ctx_outs[i_split] = gguf_init_empty();
}
gguf_add_tensor(ctx_outs[i_split], tensor);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
quantize : add '--keep-split' to quantize model into shards (#6688) * Implement '--keep-split' to quantize model into several shards * Add test script * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Split model correctly even if tensor id is out-of-order * Update llama_model_quantize_params * Fix preci failures --------- Co-authored-by: z5269887 <z5269887@unsw.edu.au> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-25 10:29:35 +00:00
// Set split info if needed
if (n_split > 1) {
for (size_t i = 0; i < ctx_outs.size(); ++i) {
gguf_set_val_u16(ctx_outs[i], ml.llm_kv(LLM_KV_SPLIT_NO).c_str(), i);
gguf_set_val_u16(ctx_outs[i], ml.llm_kv(LLM_KV_SPLIT_COUNT).c_str(), n_split);
gguf_set_val_i32(ctx_outs[i], ml.llm_kv(LLM_KV_SPLIT_TENSORS_COUNT).c_str(), ml.n_tensors);
}
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
quantize : add '--keep-split' to quantize model into shards (#6688) * Implement '--keep-split' to quantize model into several shards * Add test script * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Split model correctly even if tensor id is out-of-order * Update llama_model_quantize_params * Fix preci failures --------- Co-authored-by: z5269887 <z5269887@unsw.edu.au> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-25 10:29:35 +00:00
int cur_split = -1;
std::ofstream fout;
auto close_ofstream = [&]() {
// Write metadata and close file handler
if (fout.is_open()) {
fout.seekp(0);
std::vector<uint8_t> data(gguf_get_meta_size(ctx_outs[cur_split]));
gguf_get_meta_data(ctx_outs[cur_split], data.data());
fout.write((const char *) data.data(), data.size());
fout.close();
}
};
tests : minor bash stuff (#6902) * tests : minor bash stuff ggml-ci * llama : fix build ggml-ci * tests : fix CUR_DIR -> ROOT_DIR ggml-ci * tests : fix fname ggml-ci
2024-04-25 11:27:20 +00:00
auto new_ofstream = [&](int index) {
quantize : add '--keep-split' to quantize model into shards (#6688) * Implement '--keep-split' to quantize model into several shards * Add test script * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Split model correctly even if tensor id is out-of-order * Update llama_model_quantize_params * Fix preci failures --------- Co-authored-by: z5269887 <z5269887@unsw.edu.au> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-25 10:29:35 +00:00
cur_split = index;
GGML_ASSERT(ctx_outs[cur_split] && "Find uninitialized gguf_context");
std::string fname = fname_out;
if (params->keep_split) {
char split_path[PATH_MAX] = {0};
llama_split_path(split_path, sizeof(split_path), fname_out.c_str(), cur_split, n_split);
fname = std::string(split_path);
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
quantize : add '--keep-split' to quantize model into shards (#6688) * Implement '--keep-split' to quantize model into several shards * Add test script * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Split model correctly even if tensor id is out-of-order * Update llama_model_quantize_params * Fix preci failures --------- Co-authored-by: z5269887 <z5269887@unsw.edu.au> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-25 10:29:35 +00:00
fout = std::ofstream(fname, std::ios::binary);
fout.exceptions(std::ofstream::failbit); // fail fast on write errors
const size_t meta_size = gguf_get_meta_size(ctx_outs[cur_split]);
// placeholder for the meta data
::zeros(fout, meta_size);
};
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
const auto tn = LLM_TN(model.arch);
tests : minor bash stuff (#6902) * tests : minor bash stuff ggml-ci * llama : fix build ggml-ci * tests : fix CUR_DIR -> ROOT_DIR ggml-ci * tests : fix fname ggml-ci
2024-04-25 11:27:20 +00:00
new_ofstream(0);
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
for (int i = 0; i < ml.n_tensors; ++i) {
quantize : add '--keep-split' to quantize model into shards (#6688) * Implement '--keep-split' to quantize model into several shards * Add test script * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Split model correctly even if tensor id is out-of-order * Update llama_model_quantize_params * Fix preci failures --------- Co-authored-by: z5269887 <z5269887@unsw.edu.au> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-25 10:29:35 +00:00
auto weight = ml.get_weight(i);
struct ggml_tensor * tensor = weight->tensor;
if (weight->idx != cur_split && params->keep_split) {
close_ofstream();
new_ofstream(weight->idx);
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
const std::string name = ggml_get_name(tensor);
llama : quantize up to 31% faster on Linux and Windows with mmap (#3206) * llama : enable mmap in quantize on Linux -> 31% faster * also enable mmap on Windows --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-29 13:48:45 +00:00
if (!ml.use_mmap) {
if (read_data.size() < ggml_nbytes(tensor)) {
read_data.resize(ggml_nbytes(tensor));
}
tensor->data = read_data.data();
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
}
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
ml.load_data_for(tensor);
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
LLAMA_LOG_INFO("[%4d/%4d] %36s - [%s], type = %6s, ",
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
++idx, ml.n_tensors,
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
ggml_get_name(tensor),
llama_format_tensor_shape(tensor).c_str(),
ggml_type_name(tensor->type));
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
// This used to be a regex, but <regex> has an extreme cost to compile times.
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
bool quantize = name.rfind("weight") == name.size() - 6; // ends with 'weight'?
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
// quantize only 2D and 3D tensors (experts)
quantize &= (ggml_n_dims(tensor) >= 2);
llama : add Command R Plus support (#6491) * Add Command R Plus GGUF * Add Command R Plus GGUF * Loading works up to LayerNorm2D * Export new tensors in 1D so they are not quantized. * Fix embedding layer based on Noeda's example * Whitespace * Add line * Fix unexpected tokens on MPS. Re-add F16 fix. ((Noeda) * dranger003: Fix block index overflow in CUDA dequantizing. * Reverted blocked multiplication code as it still has issues and could affect other Llama arches * export norms as f32 * fix overflow issues during quant and other cleanup * Type convention Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * dranger003: Fix more int overflow during quant. --------- Co-authored-by: S <seast@Ss-Mac-Studio.local> Co-authored-by: S <s@example.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-09 08:16:13 +00:00
// do not quantize norm tensors
quantize &= name.find("_norm.weight") == std::string::npos;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
quantize &= params->quantize_output_tensor || name != "output.weight";
Allow quantize to only copy tensors, some other improvements (#2931) * Allow quantize tool to only copy tensors to allow repackaging models. * Slightly better logic when requantizing. * Change help message to go to `stdout`.
2023-09-01 14:02:48 +00:00
quantize &= !params->only_copy;
ggml : sync ggml (add GPT-NeoX RoPE implementation)
2023-04-20 20:32:59 +00:00
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
// do not quantize expert gating tensors
llama : fix non-quantization of expert gating tensors (#5754) This reverts a single line from #5475
2024-02-28 08:52:56 +00:00
// NOTE: can't use LLM_TN here because the layer number is not known
quantize &= name.find("ffn_gate_inp.weight") == std::string::npos;
bert : add tests + fix quantization (#5475) * llama : do not quantize pos embd and token type tensors * ci : add BERT tests ggml-ci * ci : do not do BERT tests on low-perf nodes ggml-ci
2024-02-13 11:01:29 +00:00
// do not quantize positional embeddings and token types (BERT)
quantize &= name != LLM_TN(model.arch)(LLM_TENSOR_POS_EMBD, "weight");
quantize &= name != LLM_TN(model.arch)(LLM_TENSOR_TOKEN_TYPES, "weight");
llama : add Mixtral support (#4406) * convert : support Mixtral as LLAMA arch * convert : fix n_ff typo * llama : model loading * ggml : sync latest ggml_mul_mat_id * llama : update graph to support MoE * llama : fix cur -> cur_expert * llama : first working version * llama : fix expert weighting in the FFN * ggml : ggml_get_rows support 2D indexing [n_tokens, n_experts] (cpu only) * ggml : add n_as argument to ggml_mul_mat_id * ggml : fix ggml_get_rows to take into account ne02 / ne11 * metal : add more general support for ggml_get_rows + tests * llama : add basic support for offloading moe with CUDA * metal : add/mul/div use general kernel when src1 not cont * metal : reduce the kernel launches for ggml_mul_mat_id * ggml : get_rows : support non-contiguos tensors with gaps, generalize up to 3D * ggml : update get_rows f16 and q * cuda : support non-contiguous src1 in get_rows * llama : offload missing ffn_moe_silu * metal : fix ggml_get_rows to work with non-cont src1 * metal : add indirect mat-vec kernels for all quantization types * llama : do not quantize expert gating tensors * llama : add n_expert and n_expert_used to hparams + change quants * test-backend-ops : add moe test * cuda : fix get_rows when ncols is odd * convert : determine n_ctx correctly * metal : fix ggml_mul_mat_id for F32 * test-backend-ops : make experts more evenly probable (test_moe) * test-backend-ops : cleanup, add moe test for batches * test-backend-ops : add cpy from f32 -> all types test * test-backend-ops : fix dequantize block offset * llama : fix hard-coded number of experts * test-backend-ops : simplify and disable slow tests to avoid CI timeout * test-backend-ops : disable MOE test with thread sanitizer * cuda : fix mul_mat_id with multi gpu * convert : use 1e6 rope_freq_base for mixtral * convert : fix style * convert : support safetensors format * gguf-py : bump version * metal : add cpy f16 -> f32 kernel * metal : fix binary ops for ne10 % 4 != 0 * test-backend-ops : add one more sum_rows test * ggml : do not use BLAS with ggml_mul_mat_id * convert-hf : support for mixtral-instruct (#4428) * convert : typo fix, add additional hyperparameters, use LLaMA arch for Mixtral-instruct * convert : use sentencepiece tokenizer for Mixtral-instruct * convert : make flake8 happy * metal : fix soft_max kernels ref: https://github.com/ggerganov/ggml/pull/621/commits/1914017863d2f9ab8ecc0281cc2a56d683668b92 * metal : limit kernels to not use more than the allowed threads --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Radek Pilar <github@mrkva.eu>
2023-12-13 12:04:25 +00:00
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
// do not quantize Mamba's small yet 2D weights
// NOTE: can't use LLM_TN here because the layer number is not known
quantize &= name.find("ssm_conv1d.weight") == std::string::npos;
llama : fix F16/F32 downcast + improve names (#5980)
2024-03-11 07:56:47 +00:00
quantize &= name.find("ssm_x.weight") == std::string::npos;
quantize &= name.find("ssm_dt.weight") == std::string::npos;
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
enum ggml_type new_type;
void * new_data;
size_t new_size;
Allow quantize to only copy tensors, some other improvements (#2931) * Allow quantize tool to only copy tensors to allow repackaging models. * Slightly better logic when requantizing. * Change help message to go to `stdout`.
2023-09-01 14:02:48 +00:00
if (quantize) {
llama : fix F16/F32 downcast + improve names (#5980)
2024-03-11 07:56:47 +00:00
new_type = default_type;
// get more optimal quantization type based on the tensor shape, layer, etc.
if (!params->pure && ggml_is_quantized(default_type)) {
new_type = llama_tensor_get_type(qs, new_type, tensor, ftype);
ggml : quantization refactoring (#3833) * ggml : factor all quantization code in ggml-quants ggml-ci * ggml-quants : fix Zig and Swift builds + quantize tool ggml-ci * quantize : --pure option for disabling k-quant mixtures --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com>
2023-10-29 16:32:28 +00:00
}
quantize : fix precedence of cli args (#6541)
2024-04-08 13:23:01 +00:00
if (params->token_embedding_type < GGML_TYPE_COUNT && strcmp(tensor->name, "token_embd.weight") == 0) {
IQ1_M: 1.75 bpw quantization (#6302) * iq1_m: basics * iq1_m: basics-2 * iq1_m: CUDA dequantize works Very 1st shot I get PPL = 9.76 for LLaMA-v2-7B. * iq1_m: separate shifts for each group of 8 in a block We get PPL(LLaMA-v2-7B ) = 9.2810 PPL(LLaMA-v2-13B) = 6.8105 Not bad, but slightly higher than sqrt(PPL(IQ1_S) * PPL(IQ2_XXS)) which is the expected outcome given that IQ1_M is halfway between IQ1_S and IQ2_XXS in terms of bpw. From this, we would expect PPL = 9.14 for LLaMA-v2-7B PPL = 6.63 for LLaMA-v2-13B * iq1_m: go to 3-bit scales There is slight increase in PPL, but the 0.0625 bpw reduction in size is totally worth it. We now have PPL(LLaMA-v2-7B ) = 9.4469 at 1.96 bpw PPL(LLaMA-v2-13B) = 6.8717 at 1.93 bpw PPL(LLaMA-v2-70B) = 4.8568 at 1.85 bpw * iq1_m: scalar dot product * iq1_m: AVX2 dot product * iq1_m: very slightly faster AVX2 dot product * iq1_m: ARM_NEON dot product Works, but very slow (10.5 t/s) * iq1_m: Metal - dequantize works, dot product does not * iq1_m: Metal now works About the same performance as iq1_s. * iq1_m: minor * iq1_m: checking pure iq1_m quantization It is pretty bad: PPL(LLaMA-v2-7B) = 34 if we quantize output.weight with Q4_K. * iiq1_m: slightly faster ARM_NEON dot product 10.5 t/s -> 11.65 t/s * iq1_m: faster ARM_NEON dot product 11.65 t/s -> 14.9 t/s * iq1_m: another minor ARM_NEON dot product improvement 14.9 -> 15.0 t/s * iq1_m: small PPL improvement via super-block scale adjustment After quantizing block scales redo the super-block scale fit. PPL(LLaMA-v2-7B ) = 9.3346 PPL(LLaMA-v2-13B) = 6.8419 PPL(LLaMA-v2-70B) = 4.8294 PPL(Mistral-7B ) = 8.1624 * iq1_m: adapt to CUDA refactoring * iq1_m: remove unused variable We have progressed to warnings being errors. * iq1_m: add to backend-ops tests * iq1_m: fix Windows ARM * iq1_m: use common definition of iq1m_scale_t * cuda: assert -> NO_DEVICE_CODE * iq1_M: PR comments --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-26 14:21:27 +00:00
new_type = params->token_embedding_type;
}
quantize : fix precedence of cli args (#6541)
2024-04-08 13:23:01 +00:00
if (params->output_tensor_type < GGML_TYPE_COUNT && strcmp(tensor->name, "output.weight") == 0) {
IQ1_M: 1.75 bpw quantization (#6302) * iq1_m: basics * iq1_m: basics-2 * iq1_m: CUDA dequantize works Very 1st shot I get PPL = 9.76 for LLaMA-v2-7B. * iq1_m: separate shifts for each group of 8 in a block We get PPL(LLaMA-v2-7B ) = 9.2810 PPL(LLaMA-v2-13B) = 6.8105 Not bad, but slightly higher than sqrt(PPL(IQ1_S) * PPL(IQ2_XXS)) which is the expected outcome given that IQ1_M is halfway between IQ1_S and IQ2_XXS in terms of bpw. From this, we would expect PPL = 9.14 for LLaMA-v2-7B PPL = 6.63 for LLaMA-v2-13B * iq1_m: go to 3-bit scales There is slight increase in PPL, but the 0.0625 bpw reduction in size is totally worth it. We now have PPL(LLaMA-v2-7B ) = 9.4469 at 1.96 bpw PPL(LLaMA-v2-13B) = 6.8717 at 1.93 bpw PPL(LLaMA-v2-70B) = 4.8568 at 1.85 bpw * iq1_m: scalar dot product * iq1_m: AVX2 dot product * iq1_m: very slightly faster AVX2 dot product * iq1_m: ARM_NEON dot product Works, but very slow (10.5 t/s) * iq1_m: Metal - dequantize works, dot product does not * iq1_m: Metal now works About the same performance as iq1_s. * iq1_m: minor * iq1_m: checking pure iq1_m quantization It is pretty bad: PPL(LLaMA-v2-7B) = 34 if we quantize output.weight with Q4_K. * iiq1_m: slightly faster ARM_NEON dot product 10.5 t/s -> 11.65 t/s * iq1_m: faster ARM_NEON dot product 11.65 t/s -> 14.9 t/s * iq1_m: another minor ARM_NEON dot product improvement 14.9 -> 15.0 t/s * iq1_m: small PPL improvement via super-block scale adjustment After quantizing block scales redo the super-block scale fit. PPL(LLaMA-v2-7B ) = 9.3346 PPL(LLaMA-v2-13B) = 6.8419 PPL(LLaMA-v2-70B) = 4.8294 PPL(Mistral-7B ) = 8.1624 * iq1_m: adapt to CUDA refactoring * iq1_m: remove unused variable We have progressed to warnings being errors. * iq1_m: add to backend-ops tests * iq1_m: fix Windows ARM * iq1_m: use common definition of iq1m_scale_t * cuda: assert -> NO_DEVICE_CODE * iq1_M: PR comments --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-26 14:21:27 +00:00
new_type = params->output_tensor_type;
}
ggml : quantization refactoring (#3833) * ggml : factor all quantization code in ggml-quants ggml-ci * ggml-quants : fix Zig and Swift builds + quantize tool ggml-ci * quantize : --pure option for disabling k-quant mixtures --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com>
2023-10-29 16:32:28 +00:00
Allow quantize to only copy tensors, some other improvements (#2931) * Allow quantize tool to only copy tensors to allow repackaging models. * Slightly better logic when requantizing. * Change help message to go to `stdout`.
2023-09-01 14:02:48 +00:00
// If we've decided to quantize to the same type the tensor is already
// in then there's nothing to do.
quantize = tensor->type != new_type;
}
llama : fix F16/F32 downcast + improve names (#5980)
2024-03-11 07:56:47 +00:00
Allow quantize to only copy tensors, some other improvements (#2931) * Allow quantize tool to only copy tensors to allow repackaging models. * Slightly better logic when requantizing. * Change help message to go to `stdout`.
2023-09-01 14:02:48 +00:00
if (!quantize) {
new_type = tensor->type;
new_data = tensor->data;
new_size = ggml_nbytes(tensor);
LLAMA_LOG_INFO("size = %8.3f MB\n", ggml_nbytes(tensor)/1024.0/1024.0);
} else {
llama : add Command R Plus support (#6491) * Add Command R Plus GGUF * Add Command R Plus GGUF * Loading works up to LayerNorm2D * Export new tensors in 1D so they are not quantized. * Fix embedding layer based on Noeda's example * Whitespace * Add line * Fix unexpected tokens on MPS. Re-add F16 fix. ((Noeda) * dranger003: Fix block index overflow in CUDA dequantizing. * Reverted blocked multiplication code as it still has issues and could affect other Llama arches * export norms as f32 * fix overflow issues during quant and other cleanup * Type convention Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * dranger003: Fix more int overflow during quant. --------- Co-authored-by: S <seast@Ss-Mac-Studio.local> Co-authored-by: S <s@example.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-09 08:16:13 +00:00
const int64_t nelements = ggml_nelements(tensor);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
2-bit quantizations (#4897) * imatrix: load * imatrix: WIP * imatrix: Add Q2_K quantization * imatrix: also guard against Q2_K_S quantization without importance matrix * imatrix: guard even more against low-bit quantization misuse --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:45:56 +00:00
const float * imatrix = nullptr;
if (imatrix_data) {
auto it = imatrix_data->find(tensor->name);
if (it == imatrix_data->end()) {
llama : use LLAMA_LOG_ macros for logging
2024-01-14 09:03:19 +00:00
LLAMA_LOG_INFO("\n====== %s: did not find weights for %s\n", __func__, tensor->name);
2-bit quantizations (#4897) * imatrix: load * imatrix: WIP * imatrix: Add Q2_K quantization * imatrix: also guard against Q2_K_S quantization without importance matrix * imatrix: guard even more against low-bit quantization misuse --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:45:56 +00:00
} else {
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
if (it->second.size() == (size_t)tensor->ne[0]*tensor->ne[2]) {
2-bit quantizations (#4897) * imatrix: load * imatrix: WIP * imatrix: Add Q2_K quantization * imatrix: also guard against Q2_K_S quantization without importance matrix * imatrix: guard even more against low-bit quantization misuse --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:45:56 +00:00
imatrix = it->second.data();
} else {
llama : use LLAMA_LOG_ macros for logging
2024-01-14 09:03:19 +00:00
LLAMA_LOG_INFO("\n====== %s: imatrix size %d is different from tensor size %d for %s\n", __func__,
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
int(it->second.size()), int(tensor->ne[0]*tensor->ne[2]), tensor->name);
// this can happen when quantizing an old mixtral model with split tensors with a new incompatible imatrix
// this is a significant error and it may be good idea to abort the process if this happens,
// since many people will miss the error and not realize that most of the model is being quantized without an imatrix
// tok_embd should be ignored in this case, since it always causes this warning
if (name != tn(LLM_TENSOR_TOKEN_EMBD, "weight")) {
throw std::runtime_error(format("imatrix size %d is different from tensor size %d for %s",
int(it->second.size()), int(tensor->ne[0]*tensor->ne[2]), tensor->name));
}
2-bit quantizations (#4897) * imatrix: load * imatrix: WIP * imatrix: Add Q2_K quantization * imatrix: also guard against Q2_K_S quantization without importance matrix * imatrix: guard even more against low-bit quantization misuse --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:45:56 +00:00
}
}
}
if ((new_type == GGML_TYPE_IQ2_XXS ||
new_type == GGML_TYPE_IQ2_XS ||
Adding IQ2_S and IQ2_M to complete coverage of the 2-3 bit quantization range (#5721) * Adding IQ2_S and IQ2_M as a single cumulative commit * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-26 16:28:38 +00:00
new_type == GGML_TYPE_IQ2_S ||
1.5 bit quantization (#5453) * iq1_s: WIP basics * iq1_s: CUDA is working * iq1_s: scalar CPU dot product * iq1_s: WIP AVX2 dot product - something is not right * Fix tests * Fix shadow warnings * Fix after merge with latest master * iq1_s: AVX2 finally works * iq1_s: ARM_NEON dot product. Works, but not very fast * iq1_s: better grid * iq1_s: use IQ2_XXS for attn_output At a cost of 0.04 extra bpw this gives a big improvement in PPL. * iq1_s: Metal basics Dequantize works, but not dot product * iq1_s: Metal works, but quite slow As usual, Apple Silicon does not like the code I write. * iq1_s: Tests * iq1_s: slightly faster dot product --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-02-18 16:16:55 +00:00
new_type == GGML_TYPE_IQ1_S ||
IQ1_M: 1.75 bpw quantization (#6302) * iq1_m: basics * iq1_m: basics-2 * iq1_m: CUDA dequantize works Very 1st shot I get PPL = 9.76 for LLaMA-v2-7B. * iq1_m: separate shifts for each group of 8 in a block We get PPL(LLaMA-v2-7B ) = 9.2810 PPL(LLaMA-v2-13B) = 6.8105 Not bad, but slightly higher than sqrt(PPL(IQ1_S) * PPL(IQ2_XXS)) which is the expected outcome given that IQ1_M is halfway between IQ1_S and IQ2_XXS in terms of bpw. From this, we would expect PPL = 9.14 for LLaMA-v2-7B PPL = 6.63 for LLaMA-v2-13B * iq1_m: go to 3-bit scales There is slight increase in PPL, but the 0.0625 bpw reduction in size is totally worth it. We now have PPL(LLaMA-v2-7B ) = 9.4469 at 1.96 bpw PPL(LLaMA-v2-13B) = 6.8717 at 1.93 bpw PPL(LLaMA-v2-70B) = 4.8568 at 1.85 bpw * iq1_m: scalar dot product * iq1_m: AVX2 dot product * iq1_m: very slightly faster AVX2 dot product * iq1_m: ARM_NEON dot product Works, but very slow (10.5 t/s) * iq1_m: Metal - dequantize works, dot product does not * iq1_m: Metal now works About the same performance as iq1_s. * iq1_m: minor * iq1_m: checking pure iq1_m quantization It is pretty bad: PPL(LLaMA-v2-7B) = 34 if we quantize output.weight with Q4_K. * iiq1_m: slightly faster ARM_NEON dot product 10.5 t/s -> 11.65 t/s * iq1_m: faster ARM_NEON dot product 11.65 t/s -> 14.9 t/s * iq1_m: another minor ARM_NEON dot product improvement 14.9 -> 15.0 t/s * iq1_m: small PPL improvement via super-block scale adjustment After quantizing block scales redo the super-block scale fit. PPL(LLaMA-v2-7B ) = 9.3346 PPL(LLaMA-v2-13B) = 6.8419 PPL(LLaMA-v2-70B) = 4.8294 PPL(Mistral-7B ) = 8.1624 * iq1_m: adapt to CUDA refactoring * iq1_m: remove unused variable We have progressed to warnings being errors. * iq1_m: add to backend-ops tests * iq1_m: fix Windows ARM * iq1_m: use common definition of iq1m_scale_t * cuda: assert -> NO_DEVICE_CODE * iq1_M: PR comments --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-26 14:21:27 +00:00
(new_type == GGML_TYPE_IQ1_M && strcmp(tensor->name, "token_embd.weight") && strcmp(tensor->name, "output.weight")) ||
2-bit quantizations (#4897) * imatrix: load * imatrix: WIP * imatrix: Add Q2_K quantization * imatrix: also guard against Q2_K_S quantization without importance matrix * imatrix: guard even more against low-bit quantization misuse --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:45:56 +00:00
(new_type == GGML_TYPE_Q2_K && params->ftype == LLAMA_FTYPE_MOSTLY_Q2_K_S && strcmp(tensor->name, "token_embd.weight") != 0)) && !imatrix) {
llama : use LLAMA_LOG_ macros for logging
2024-01-14 09:03:19 +00:00
LLAMA_LOG_ERROR("\n\n============================================================\n");
LLAMA_LOG_ERROR("Missing importance matrix for tensor %s in a very low-bit quantization\n", tensor->name);
LLAMA_LOG_ERROR("The result will be garbage, so bailing out\n");
LLAMA_LOG_ERROR("============================================================\n\n");
2-bit quantizations (#4897) * imatrix: load * imatrix: WIP * imatrix: Add Q2_K quantization * imatrix: also guard against Q2_K_S quantization without importance matrix * imatrix: guard even more against low-bit quantization misuse --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:45:56 +00:00
throw std::runtime_error(format("Missing importance matrix for tensor %s in a very low-bit quantization", tensor->name));
}
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
float * f32_data;
llama : support requantizing models instead of only allowing quantization from 16/32bit (#1691) * Add support for quantizing already quantized models * Threaded dequantizing and f16 to f32 conversion * Clean up thread blocks with spares calculation a bit * Use std::runtime_error exceptions.
2023-06-10 07:59:17 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
if (tensor->type == GGML_TYPE_F32) {
f32_data = (float *) tensor->data;
} else if (ggml_is_quantized(tensor->type) && !params->allow_requantize) {
throw std::runtime_error(format("requantizing from type %s is disabled", ggml_type_name(tensor->type)));
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
} else {
llama : refactor internal quantization functions (#5830)
2024-03-02 14:19:09 +00:00
llama_tensor_dequantize_internal(tensor, f32_conv_buf, workers, nelements, nthread);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
f32_data = (float *) f32_conv_buf.data();
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
llama : fix F16/F32 downcast + improve names (#5980)
2024-03-11 07:56:47 +00:00
LLAMA_LOG_INFO("converting to %s .. ", ggml_type_name(new_type));
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
fflush(stdout);
llama : add Command R Plus support (#6491) * Add Command R Plus GGUF * Add Command R Plus GGUF * Loading works up to LayerNorm2D * Export new tensors in 1D so they are not quantized. * Fix embedding layer based on Noeda's example * Whitespace * Add line * Fix unexpected tokens on MPS. Re-add F16 fix. ((Noeda) * dranger003: Fix block index overflow in CUDA dequantizing. * Reverted blocked multiplication code as it still has issues and could affect other Llama arches * export norms as f32 * fix overflow issues during quant and other cleanup * Type convention Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * dranger003: Fix more int overflow during quant. --------- Co-authored-by: S <seast@Ss-Mac-Studio.local> Co-authored-by: S <s@example.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-09 08:16:13 +00:00
if (work.size() < (size_t)nelements * 4) {
llama : make quantize example up to 2.7x faster (#3115)
2023-09-15 01:09:53 +00:00
work.resize(nelements * 4); // upper bound on size
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
new_data = work.data();
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
llama : add Command R Plus support (#6491) * Add Command R Plus GGUF * Add Command R Plus GGUF * Loading works up to LayerNorm2D * Export new tensors in 1D so they are not quantized. * Fix embedding layer based on Noeda's example * Whitespace * Add line * Fix unexpected tokens on MPS. Re-add F16 fix. ((Noeda) * dranger003: Fix block index overflow in CUDA dequantizing. * Reverted blocked multiplication code as it still has issues and could affect other Llama arches * export norms as f32 * fix overflow issues during quant and other cleanup * Type convention Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * dranger003: Fix more int overflow during quant. --------- Co-authored-by: S <seast@Ss-Mac-Studio.local> Co-authored-by: S <s@example.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-09 08:16:13 +00:00
const int64_t n_per_row = tensor->ne[0];
const int64_t nrows = tensor->ne[1];
2-bit quantizations (#4897) * imatrix: load * imatrix: WIP * imatrix: Add Q2_K quantization * imatrix: also guard against Q2_K_S quantization without importance matrix * imatrix: guard even more against low-bit quantization misuse --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:45:56 +00:00
llama : add Command R Plus support (#6491) * Add Command R Plus GGUF * Add Command R Plus GGUF * Loading works up to LayerNorm2D * Export new tensors in 1D so they are not quantized. * Fix embedding layer based on Noeda's example * Whitespace * Add line * Fix unexpected tokens on MPS. Re-add F16 fix. ((Noeda) * dranger003: Fix block index overflow in CUDA dequantizing. * Reverted blocked multiplication code as it still has issues and could affect other Llama arches * export norms as f32 * fix overflow issues during quant and other cleanup * Type convention Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * dranger003: Fix more int overflow during quant. --------- Co-authored-by: S <seast@Ss-Mac-Studio.local> Co-authored-by: S <s@example.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-09 08:16:13 +00:00
static const int64_t min_chunk_size = 32 * 512;
const int64_t chunk_size = n_per_row >= min_chunk_size ? n_per_row : n_per_row * ((min_chunk_size + n_per_row - 1)/n_per_row);
2-bit quantizations (#4897) * imatrix: load * imatrix: WIP * imatrix: Add Q2_K quantization * imatrix: also guard against Q2_K_S quantization without importance matrix * imatrix: guard even more against low-bit quantization misuse --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:45:56 +00:00
llama : add Command R Plus support (#6491) * Add Command R Plus GGUF * Add Command R Plus GGUF * Loading works up to LayerNorm2D * Export new tensors in 1D so they are not quantized. * Fix embedding layer based on Noeda's example * Whitespace * Add line * Fix unexpected tokens on MPS. Re-add F16 fix. ((Noeda) * dranger003: Fix block index overflow in CUDA dequantizing. * Reverted blocked multiplication code as it still has issues and could affect other Llama arches * export norms as f32 * fix overflow issues during quant and other cleanup * Type convention Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * dranger003: Fix more int overflow during quant. --------- Co-authored-by: S <seast@Ss-Mac-Studio.local> Co-authored-by: S <s@example.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-09 08:16:13 +00:00
const int64_t nelements_matrix = tensor->ne[0] * tensor->ne[1];
const int64_t nchunk = (nelements_matrix + chunk_size - 1)/chunk_size;
const int64_t nthread_use = nthread > 1 ? std::max((int64_t)1, std::min((int64_t)nthread, nchunk)) : 1;
Make loading weights 10-100x faster This is a breaking change that's going to give you three benefits: 1. Your inference commands should load 100x faster 2. You may be able to safely load models 2x larger 3. You can run many concurrent inference processes This was accomplished by changing the file format so we can mmap() weights directly into memory without having to read() or copy them thereby ensuring the kernel can make its file cache pages directly accessible to our inference processes; and secondly, that the file cache pages are much less likely to get evicted (which would force loads to hit disk) because they're no longer competing with memory pages that were needlessly created by gigabytes of standard i/o. The new file format supports single-file models like LLaMA 7b, and it also supports multi-file models like LLaMA 13B. Our Python tool now merges the foo.1, foo.2, etc. files back into a single file so that the C++ code which maps it doesn't need to reshape data every time. That's made llama.cpp so much simpler. Much of its load code has now been deleted. Furthermore, this change ensures that tensors are aligned properly on a 32-byte boundary. That opens the door to seeing if we can get additional performance gains on some microprocessors, by using ops that require memory alignment. Lastly note that both POSIX and the Windows platform are supported Fixes #91
2023-03-29 20:51:37 +00:00
ggml : mul_mat_id use the same tensor for all the experts (#6387) * ggml : update mul_mat_id to use the same tensor for all the experts * update cuda * minor * update metal * update test-backend-ops * fix cuda * Update ggml-metal.m Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * update convert.py * update convert-hf-to-gguf.py * update convert.py for mixtral hf models * Update convert-hf-to-gguf.py Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * cuda : support non-pow-2 number of experts * allow quantize to work for split and merged experts models in the same way * cleanup + disable mmap automatically with split tensors models * update imatrix * test-backend-ops : test qwen argsort * update grok model loading * llama : add merged experts tensors to the grok tensor map * minor * gguf : bump version * fix quantizing of merged experts * convert-hf-to-gguf.py : update grok (untested) * make linter happy * cuda/argsort : use shared memory instead of pool memory * convert : fix grok tensor names * metal : add support for non-pow-2 argsort * llama : more loader cleanup, better error checking * cuda : fix warning * llama : still use mmap for loading old models, but copy the data to a host buffer * add review note * llama : remove ffn tensor counting + add sanity check ggml-ci * convert : fix handling of n_experts == None ggml-ci * imatrix : fix ncall counters * llama : produce error if imatrix size does not match * quantize : terminate on errors + trace logs ggml-ci * metal : pad shared memory to 16 bytes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-03 13:07:05 +00:00
// quantize each expert separately since they have different importance matrices
new_size = 0;
for (int64_t i03 = 0; i03 < tensor->ne[2]; ++i03) {
const float * f32_data_03 = f32_data + i03 * nelements_matrix;
void * new_data_03 = (char *)new_data + ggml_row_size(new_type, n_per_row) * i03 * nrows;
const float * imatrix_03 = imatrix ? imatrix + i03 * n_per_row : nullptr;
new_size += llama_tensor_quantize_internal(new_type, f32_data_03, new_data_03, chunk_size, nrows, n_per_row, imatrix_03, workers, nthread_use);
}
ggml : remove old quantization functions (#5942) * ggml : remove old quantization functions ggml-ci * ggml : simplify ggml_quantize_chunk ggml-ci * ggml : restrict correctness ggml-ci * ggml : remove hist data from the quantization API ggml-ci * tests : remove hist usage in test-backend-ops ggml-ci * vulkan : remove hist and fix typo
2024-03-09 13:53:59 +00:00
LLAMA_LOG_INFO("size = %8.2f MiB -> %8.2f MiB\n", ggml_nbytes(tensor)/1024.0/1024.0, new_size/1024.0/1024.0);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
total_size_org += ggml_nbytes(tensor);
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
total_size_new += new_size;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
// update the gguf meta data as we go
quantize : add '--keep-split' to quantize model into shards (#6688) * Implement '--keep-split' to quantize model into several shards * Add test script * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Split model correctly even if tensor id is out-of-order * Update llama_model_quantize_params * Fix preci failures --------- Co-authored-by: z5269887 <z5269887@unsw.edu.au> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-25 10:29:35 +00:00
gguf_set_tensor_type(ctx_outs[cur_split], name.c_str(), new_type);
gguf_set_tensor_data(ctx_outs[cur_split], name.c_str(), new_data, new_size);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
// write tensor data + padding
fout.write((const char *) new_data, new_size);
zeros(fout, GGML_PAD(new_size, align) - new_size);
}
quantize : add '--keep-split' to quantize model into shards (#6688) * Implement '--keep-split' to quantize model into several shards * Add test script * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Split model correctly even if tensor id is out-of-order * Update llama_model_quantize_params * Fix preci failures --------- Co-authored-by: z5269887 <z5269887@unsw.edu.au> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-25 10:29:35 +00:00
close_ofstream();
for (auto & c:ctx_outs) {
gguf_free(c);
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
LLAMA_LOG_INFO("%s: model size = %8.2f MB\n", __func__, total_size_org/1024.0/1024.0);
LLAMA_LOG_INFO("%s: quant size = %8.2f MB\n", __func__, total_size_new/1024.0/1024.0);
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
if (qs.n_fallback > 0) {
llama : fix F16/F32 downcast + improve names (#5980)
2024-03-11 07:56:47 +00:00
LLAMA_LOG_WARN("%s: WARNING: %d of %d tensor(s) required fallback quantization\n",
llama : allow quantizing k-quants to fall back when tensor size incompatible (#3747) * Allow quantizing k-quants to fall back when tensor size incompatible * quantizing: Add warning when tensors were incompatible with k-quants Clean up k-quants state passing a bit
2023-10-28 11:54:24 +00:00
__func__, qs.n_fallback, qs.n_k_quantized + qs.n_fallback);
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
check C++ code with -Wmissing-declarations (#3184)
2023-09-15 19:38:27 +00:00
static int llama_apply_lora_from_file_internal(
train : finetune LORA (#2632) * fix track_max_mem in forward_batch_wo_cache_flash_attn_train * remove unnecessary Adam(W) optimizer tensors. reduces optimizer memory overhead from 7*modelsize to 2*modelsize. additionally allows to optimize models with more than 2^31 parameters by replacing int with int64_t. bumps training checkpoint file version, but old checkpoints can still be read. new version with less tensors is saved. * add gradient clipping to AdamW * Fix reset of unused g->nodes and g->grads to NULL * implement gradient checkpointing for training reduces memory overhead from O(n_layer) to O(sqrt(n_layer)) as explained in readme of https://github.com/cybertronai/gradient-checkpointing * remove unused compute buffer 3 * add and use function ggml_build_backward_expand to avoid stack overflows with large maximum number of nodes GGML_API void ggml_build_backward_expand(struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, bool keep); * change AdamW decay parameter to work like the torch AdamW decay parameter It is now relative to Adam learning rate `alpha*sched`. Before that it was relative to `sched` only. `alpha` being the maximum learning rate and `sched` being a scaling parameter in [0..1] * change default AdamW weight decay parameter used in training to 0.1 as used in nanoGPT * change default AdamW weight decay parameter defined in ggml to 0.0, making Adam default instead of AdamW btw: the default weight decay parameter for torch.optim.AdamW is 0.01 * bug fixes for cross entropy loss ggml_cross_entropy_loss: sums where not correctly added in workload of each thread ggml_cross_entropy_loss_back: simplify backward process, reducing numerical issues guard usage of exp f16 lookup in cross entropy by #define GGML_CROSS_ENTROPY_EXP_FP16 cross entropy loss is only used once during training, but it is quite sensitive to numerical errors introduced by exp-f16-lookup. so exp-f16-lookup for cross entropy loss is disabled by default, trading better gradients for very slightly worse runtime performance. * fix test-grad0 for cross_entropy_loss the second argument to cross_entropy_loss must sum up to 1 for each row * fix test-grad0 for soft_max dont use only sum as aggregation, because sum of softmax is always 1 -> finite differences should not work instead use sum(log(soft_max()*(1-eps)+eps)); use eps to avoid log(0) * improve finite differences of test-grad0 by using double instead of float * change cross_entropy_loss to output average over all rows this helps keeping the loss and gradients in a sane range * improve gradient checkpointing sqrt(n_layers) is only the best checkpoint step when mem size of checkpoints and mem size of layers are equal. since layers require more memory than the single-tensor-checkpoint we use, the optimal values are compute different: ``` given: n, u, v objective: minimize(a*u+b*v) where a*b=n, a>0, b>0 b=n/a minimize(a*u+v*n/a) diff(a*u+v*n/a, a) = u - (v*n/a)/a diff(a*u+v*n/a, a) == 0 u - (v*n/a)/a == 0 u == v*n/(a*a) u*a*a = v*n a*a = v*n/u a = sqrt(n*v/u) ``` this change results in more checkpoints, requiring less layers to store between checkpoints, overall improving memory usage. * disable gradient checkpointing debug output * llama : fix rope usage in train-text-from-scratch after ChatGLM change * add more training parameters: --enable-restart N Only for Adam optimizer. Enable restarts of cos-decay --disable-restart N Only for Adam optimizer. Disable restarts of cos-decay --opt-past N Number of optimization iterations to track for delta convergence test. Disabled when zero. --opt-delta N Maximum delta for delta convergence test. Disabled when <= zero. --opt-max-no-improvement N Maximum number of optimization iterations with no improvement. Disabled when <= zero. --adam-epsf N AdamW epsilon for convergence test. Disabled when <= zero. --adam-min-alpha N Adam minimum learning rate alpha, usually 0.1 * alpha * replace memcpy with reshape operation so that the graph is not cut at the input this makes it possible to store other values into the input tensor and then simply recompute the graph without rebuilding it * remove unused function argument from get_example_targets_batch * measure and print total training time * add optimization callback to ggml_opt_resume_g this callback is called before each iteration with custom data and pointer to learning schedule parameter (only used in Adam(W)). can be used for dynamic learning schedule and setting input data for batches before each iteration * use optimization callback in training allows dynamic learning schedule and different batch data for each iteration without relying on low n_iter and high n_examples parameters reduces runtime by avoiding restart of optimization function and improves training convergence by providing a different batch for each iteration * add minimum number of tensor dimensions to apply weight decay (default 2) this allows to not apply weight decay to bias parameters * rename training parameter cos-decay-alpha to cos-decay-min and clarify that adam-min-alpha also applies to warmup * fix increase of model.train_samples and model.train_tokens now that each optimizer iteration gets its own batch we need to multiply by number of opt iterations * change sampling parameters for prediction after training to defaults of common.h and clarify what is context for prediction and what are generated tokens * tighten abs error bounds for cross_entropy_loss in test-grad0 * add conditional compilation of using F16 exp in flash attention uncomment `// #define GGML_FLASH_ATTN_EXP_FP16` to enable usage of f16 exp in flash attention * tighten abs error bounds for flash_attn in test-grad0 * tighten abs error bounds for sqrt in test-grad0 * remove out-commented vectorized code of opt_adam the vectorized code might be bit faster for low number of parameters, but it had a big memory usage overhead * ggml : update ggml_rms_norm_back with configurable eps * llama training : fix ggml_rms_norm_back calls to pass configurable eps * remove trailing whitespace * add train function using automatic gradient checkpointing backward pass and allocator * in train function replace add_inplace by regular add because using add_inplace seems to result in different gradients * don't use allocate hash_map on context because the context has no_alloc=True when using memory allocator resulting in NULL data pointers * correctly clone reshape and permute operations by also cloning tensor->nb values * fix variable name and add missing type cast * terminate recursive tensor cloning when reaching tensor without src tensors * correctly clone view tensors by setting data pointers without this the checkpointing would only work when being used together with memory allocator * fix variable names * swap arguments to commutative ops to be the same as in `forward_batch_wo_cache_flash_attn` * add input tensors as checkpoints so that recursive tensor cloning of gradient checkpointing terminates on input tensors * fix variable name and add missing boolean negation * make sure some tensors are not reallocated by inserting new temporary nodes depending on them: output and parameter gradient tensors need to be available at the end of the graph execution parameter gradient tensors also need to be available before the graph execution because they are set to zero before each optimizer iteration checkpoint tensors are allocated all together to reduce memory allocator fragmentation afterwards, in addition to the temporary nodes, we also need to reset the temporary leafs * fix ASSERT to work with zero layers * add training options whether to use allocator and/or unified training function * integrate unified training function which may use memory allocator the unified training function also supports arguments whether to use flash attention and/or gradient checkpointing * format name of cloned tensors with " (clone)" suffix * set names for tensors in unified train function for easier debugging * allocate graph on context using ggml_new_graph * remove handwritten training functions * remove unused training parameters "use_scratch" and "use_unified" * remove trailing whitespace * remove unused train params: mem_compute1_gb & mem_compute2_gb mem_compute_gb is used for compute when automatic memory allocator is not enabled, otherwise it can be very small to only hold the tensor definitions mem_compute0_gb is used for automatic memory allocator (as long as measurement of max required size is not implemented) * remove unused forward_batch function * add debug asserts in ggml_allocr_alloc to some common pitfalls when using this function directly * only use ggml_allocr_alloc when tensor has NULL data and is no view * fix test when to create temporary backward graph temporary backward graph is only necessary when using checkpointing * fix memory "leak" in optimizers each iteration a new cplan with new memory for work data was allocated. now cplan creation only happens at the start of optimization, with each iteration reusing the cplan and its work data. * reverse order of for loop in ggml_build_backward_expand to save memory when using gradient checkpointing and allocator with this loop order gradient checkpointing with allocator on 16 layer model saves 13% memory; 2 layer memory it saves 2% memory. the computation results are the same * add API functions to access llama model tensors * add stub example for finetuning, based on train-text-from-scratch * move and remove code * add API functions to access remaining model parameters: mult, head and rot * first draft for LORA finetune training * remove const model and layer arguments in API functions for accessing model tensors * bug fixes to make finetune compile automatic allocator does not work yet * add debug prints for training memory improvements * fix names of lora tensors * avoid stack overflow resulting from big ggml_cgraph replace stack allocation and ggml_build_forward by ggml_new_graph in combination with ggml_build_forward_expand * replace llama API functions to get model tensors by one function to get model tensor by name LLAMA_API struct ggml_tensor * llama_get_model_tensor(struct llama_model * model, const char * name); * remove unused call to not existing llama_get_layer_from_model * implement ggml_compute_forward_out_prod_q_f32 * remove trailing whitespace * add lora finetune support on quantized base model tensors * add ggml_add_cast API function this function works like ggml_add, but accepts a data type for the resulting tensor. only supported for quantized src0 input. * use ggml_add_cast in finetuning lora-applied weights will now have data type F32, which improves gradients when finetuning quantized base models * bug fix: actually use result type passed to ggml_add_cast * make sure base model tensors data cannot be used in viewable operations memory allocator would try to make lora application inplace on base model tensors. since those are memory mapped this will result in memory access violations * fix bug in ggml_out_prod which resulted in wrong n_dims of result tensors * avoid keeping in memory ALL of the gradients The problem here stems from ggml_graph_reset. This function is called in the optimization function, before each graph computation, to reset the gradients to zero. This required a unique memory slot for each gradient: allocating memory from a previosly freed memory location might lead to non-zero input gradients. During ggml_compute_backward the gradients are build stepwise by adding or substracting new values, starting from a OP_NONE tensor which needs to contain zero-values. This requires the graph reset. To avoid this I now remember in ggml_build_backward_expand the original OP_NONE gradient tensors in a hash table, which is passed to ggml_compute_backward. There instead of using add (or sub or similar) I test whether the existing gradient to be changed is a zero-valued-tensor by looking up its existence in the hash table. When it is such a zero-tensor it will not be modified, but replaced by the value to be added, otherwise the regular add (not inplace, allocator will take care of this) will be used. This way none of those zero-tensor values will be necessary in the final backward graph and more importantly they won't need a unique memory slot, just to make them zero. * remove trailing whitespace * remove debug prints and function to compute tensor data hash * improve optimization iteration prints * adjust maximal values to support finetuning 3B models * change default finetune params lora_r and lora_alpha to match the n_rank parameters of 4 * bug fix: make sure finetune input gradient is allocated at begin and kept until end * remove unnecessary src tensor from ggml_get_rows_back we don't need data of src[2] for computation, only to setup the correct output shape. remove dependency on src[2], so that allocator can work more freely. the computational graph is still completely determined, because the output shape is naturally included. this is similar to how ggml_reshape does it. * remove unnecessary src tensor from ggml_repeat & ggml_repeat_back we don't need data of src[1] for computation, only to setup the correct output shape. remove dependency on src[1], so that allocator can work more freely. the computational graph is still completely determined, because the output shape is naturally included * resolve todo allocator will only make it inplace when they are of the same type * mixing multiple LORA adapters is now possible pass more than one '--lora FNAME' argument to apply more than one LORA. use '--lora-scaled FNAME S' when you want to specify a user-defined scale for an adapter. * add option to save finetune output every N iterations * also save latest finetune output with ITERATION="LATEST" and print where files are saved saving with LATEST makes it easier to resume training from the latest checkpoint the string "LATEST" can be configured with command line option "--fn-latest STR" * update checkpoint train stats before saving via "--save-every" * add command line option `--rank-wo N` for rank of wo tensor * update finetune README * fix dump_non_result_info_yaml to output multiple lora adapters * bug fix: replace GGML_TYPE_SIZE[t] by ggml_type_size(t) * replace llama_n_mult by llama_n_ff * finetune bug fixes to compile with merged in code from master * remove prediction related code to reduce duplicated code with main use main instead * reduce large memory overhead in train-text-from-scratch all gradients had to be pinned so that graph_reset works correctly. this is no longer necessary with the changes to ggml_compute_backward introduced in this PR. * add comment explaining why finetune checkpoints are allocated in one block * make default value of float member a float literal * handle rms_norm and rope parameters the same as in train-text-from-scratch * remove unused code * remove vocab related code as it is unnecessary * add LLM_KV_TRAINING_TYPE to train-text-from-scratch checkpoints so that they can be differentiated from lora finetune checkpoints * add gguf constants and load/save functions from train-text-from-scratch * add load & save lora finetune checkpoints via gguf * add python script to convert old finetune checkpoint files to gguf * remove old checkpoint save & load code * remove code to print data checksums which was used to verify correctness of new gguf code * omit tokenization when training is disabled, only save llama lora adapter training can be disabled by passing '-n 0' to finetune * remove trailing whitespace * update README.md * implement ggml_compute_forward_repeat_f16 * avoid stack overflow of large cgraphs in test-grad0 * add ggml API functions ggml_unravel_index, ggml_get_i32_nd and its analogs for set and for f32 ggml_get_i32_1d, ggml_set_i32_1d, ggml_get_f32_1d, ggml_set_f32_1d now support non-contiguous tensors. in case of non-contiguous tensor, the 1d index is unraveled into a multi index using ggml_unravel_index to be passed to '_nd' function equivalent. this fixes a bug in test-grad0 which happens due to ggml_build_backward not building purely contiguous tensors anymore * increase test-grad0 context mem size to accommodate for bigger cgraph * add sanity check to ggml_compute_backward, asserting the correct shape of gradients * fix ggml_acc_or_set to return tensor of correct shape * remove unused 'inplace' argument from ggml_compute_backward function inplace operations to add gradients are no longer created by ggml_compute_backward use allocator to automatically make inplace operations * add missing argument 'int i0' to ggml_get_i32_nd & ggml_set_i32_nd header declarations * fix error message in ggml_allocr_alloc to display actual max_avail * fix check_gradient ggml_build_backward_expand was previously replaced by ggml_build_backward, but the assignment of forward graph to backward graph missing * use tensor->view_src instead of ggml_is_view and get_view_source * move gradient checkpointing code into ggml, new API function: // build gradient checkpointing backward graph gb for gf using provided checkpoints // gb_tmp will contain original backward graph with rewritten backward process nodes, // but without the second forward pass nodes. GGML_API void ggml_build_backward_gradient_checkpointing( struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, struct ggml_cgraph * gb_tmp, struct ggml_tensor * * checkpoints, int n_checkpoints); * replace custom data getters and setters by ggml functions * train-text-from-scratch can train (full finetune) gguf models just pass the gguf model via `--checkpoint-in FN`. after this, to continue training, pass the generated checkpoint instead of the original gguf model. tested with smaller models, bigger models may exceed available memory. use (LORA) finetune for those. * remove trailing whitespace * add option to save train-text-from-scratch output every N iterations * update README.md * fix warnings * fix warnings * remove finetune option to disable allocator the allocator should always be used. by making sure that it is always used it gets easier to implement automatic memory requirements computation * add tensor checkpoints only when gradient checkpointing is enabled * initialize opt ggml context if none was provided * add ggml-alloc API function 'ggml_allocr_max_size' to get max size of alloc GGML_API size_t ggml_allocr_max_size(struct ggml_allocr * alloc); * finetune: automatically allocate all memory and changes to command line options remove '--n_examples N' parameter, as it no longer makes sense to call optimization process multiple times in a loop. add '--only_write_lora' command line option: will skip tokenization and training, to only write a llama.cpp comptabile LORA adapter. remove memory buffer related command line options. improve iteration console output. * add finetune to Makefile * update README.md * print time per iteration and estimate remaining time * increase measured alloc size by tensor_alignment ggml_allocr_reset will reduce the given size by up to tensor_alignment-1 * fix README.md * add some more allocator debug prints * bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue * revert last commit "bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue" "alloc was freeing an externally allocated tensor, because it calculated the end of allocator memory as alloc->data + alloc->max_size instead of alloc->data + alloc->size." This is intentional to reduce the risk of freeing external tensors when measuring. Unless max_size is not properly calculated, I don't see why this is an issue. * remove unnecessary "0x" before "%p" output * move measurement memory segment to upper region of the address space * update README.md * fix printf format warnings * add missing gguf_free in load_checkpoint_lora_file * load default rms_norm and rope parameters from base model * add gradient accumulation specify number accumulation steps with '--grad-acc N'. this will simulate a bigger batch size of grad_acc*batch. * fix tracking of train_samples and train_tokens * build : fix compile warnings * ggml : fix L-BFGS linesearch loop * improve finetune time measurement fix printf warnings on system where int64_t is (long int). change time datatypes to double because values get big with long training times. exclude file saving from time measurement. converge faster to actual time per iteration by removing very small first duration before first iteration was performed. fix bug in output of total training time, the reported value was 1000 times to small. * specify default lora rank with '--lora-r N' '--lora-r N' will specify default rank for all tensors '--rank-wq N', etc. will override this default rank for specific tensor types. * fix gradient accumulation bug where the same batch was used for each microstep * fix gradient accumulation bug where the same batch was used for each microstep * support grouped-query-attention in ggml_flash_attn and ggml_flash_attn_back k and v can now be repeated in q along ne[2] in forward pass just use modulo to compute k and v indices, like ik2 = iq2 % nek2. in backard pass this won't work as easy, because multiple threads will compete to accumulate to the same k->grad[:,ik1,ik2,ik3] and v->grad[:,iv1,iv2,iv3]. so we change the parallelization over q rows to be over k rows. this ensures non-overlapping (ik2,ik3) across threads. in each thread we then iterate over the number of repetitions of k/v in q to compute iq2 as iq2 = ik2 + irep*nek2. since ne2 is not the same for q,k and v we also change how the gradients are concatenated into the result tensor. additionally the offsets of gradq, gradk and gradv in the result tensor are now memory aligned. we also simplify the compute_backward part of flash_attn to use ggml_reshape instead of switching over the number of dimensions. this needs a small change to ggml_reshape, removing the assertion of second argument to be contiguous. since only the shape (ne) of the second reshape argument is of relevance, its memory layout (nb) is irrelevant -> it can very well be non-contiguous. change test-grad0 to also test for repeated k/v in q. this changes the rng and now results in small gradient differences in softmax. these solely come from using f16 exp table lookup in forward softmax: when temporarily changing softmax to use actual exp function, the reported gradient differences go away. gradient differences coming solely from f16 table lookup are acceptable. added a note to explain this. * add llama API functions to get grouped-query-attention n_head parameter 'n_head_kv'. * fix finetune to support grouped-query-attention (using flash-attention) note: ggml changes to ggml_out_prod are necessary to support grouped-query-attention without flash-attention. * support broadcastable a in out_prod(a, b) and backward pass of broadcasting mul_mat(a, b) * test broadcasting mul_mat backward pass * decouple random number generator of each operation test when changing one test the rng of others tests is not influenced anymore * add comment briefly describing what ggml_repeat_back does * simplify broadcasting mul_mat backward using ggml_repeat_back * add cgraph evaluation order member and corresponding enum type this controls in which order ggml_build_forward visits source nodes. by default the nodes are visited left to right, i.e. src[0] first. in some cases it is beneficial for ggml-alloc to visit in a different order. two possible orders are supported: left-to-right (src[0] first) and right-to-left (src[0] last). * measure max compute size for each cgraph eval order and use best order this can bring huge memory savings: e.g. codellama-34b with n_ctx=64, n_batch=1 goes from 92927.8mb down to 4627.6 MB * remove unused command line options * add sample start patterns and options to force new or by default resume last shuffling * update shuffle rng state on reshuffle * exclude known zero values from computations in flash_attn_f32 & flash_attn_back_f32 * remove probably unnecessary exception type flags from stringstream * pass correct max number of tokens to llama_tokenize * account for possible leading whitespace that will be added by tokenizer e.g. '\t' will be tokenized by llama spm tokenizer to [29871, 12] * use unrolled vec_mad in out_prod y is vec_mad result vec. x is vec_mad input vec. v is vec_mad input scalar. ggml_vec_mad_f32_unroll will internally loop over x and v with same y. GGML_VEC_MAD_UNROLL is by default defined to 32. This value is empirical optimized using performance test runs of out-prod in openllama-3b finetune with 256 context length and batch size 1. It gives 23% performance boost for out_prod. Full measurements of out-prod runtime in ms: unroll_xv unroll_yv 1 67014.643 87826.469 2 77117.552 89077.656 4 72091.311 109121.657 8 61077.543 88678.334 16 56914.67 79514.947 24 59024.595 84350.254 28 55952.446 83368.73 32 51476.658 85177.745 36 55973.792 84659.92 40 55139.616 93844.738 48 60736.392 93330.267 64 99856.878 116994.99 Second column is when unrollying yv instead of xv * set lora_alpha to value of lora_r if it is not set via command line otherwise only changing lora_r will change scaling of lora adapter used in prediction * reshuffle original sample order instead of the previous shuffled order otherwise resumed reshuffle will not result in same sample order * block tiling for out-prod inspired by mul-mat block sizes are empirically optimized roughly doubles the flops of out-prod * exclude some more known zero values from computations in flash_attn_f32 & flash_attn_back_f32 * add static keywords * remove outcommented old code * update train-text-from-scratch with tokenization, sample selection and shuffling from finetune * remove lbfgs related train parameters * move common train functions into common/train.[h|cpp] * move train state into struct train_state * move train data saving code into callback to unify code of opt_callback train_params are still different in finetune and train-text-from-scratch, so it can't yet be moved to train.h|cpp * move common train params into common/train * move common opt_callback into common/train * fix consume_common_train_arg * save and load head_count_kv in lora checkpoints * increase train_samples by used_samples instead of number of batches on batch can contain more than one sample when option "fill_with_next_samples" is used * fix usage of llama_tokenize * remove static from process_escape since we need it exposed in header * fix code formating of long function declarations * fix condition in load_train_state_gguf * use die("msg") instead of replace GGML_ASSERT(!"msg") or throw std::runtime_error("msg") * fix saving and loading of training type * remove terminating '\0' from tokenization (llama_tokenize is now passed the string length instead of relying on terminating '\0') * fix compile warnings * fix compile warnings * use new/delete for train_state instead of malloc/free using malloc may result in seg faults when trying to assign string fields * assert that sample_count > 0, avoiding division by zero * fix frand to return value in interval [0,1) * add train option "--sample-random-offsets" Use samples beginning at random offsets. The offset is only applied to the first sample in each batch context window. Together with "--fill-with-next-samples" this may help for training endless text generation. For example given a dataset containing samples "abcd", "ABCD", "0123". With context size of 8 and options "--fill-with-next-samples", "--no-separate-with-eos", "--no-separate-with-bos", the context windows of batches could only be filled with "abcdABCD", "ABCDabcd", "0123abcd", etc. With "--sample-random-offsets" it can also be filled with "23abcdAB", "bcd0123A", etc. * deduplicate code into function * remove n_rot hparam, as it must always be hparam.n_embd_head() * align code * assert correct base model tensor shapes * move some params from lora hparams into model hparams and load model params from gguf this equalizes the model definition in finetune and text-from-scratch and removes the need for additional llama api functions to get model parameters * remove now unnecessary llama API functions to get model params that where added by this PR * train-text-from-scratch: automatically allocate model tensors, remove option '--mem-model N' * train-text-from-scratch: automatically allocate opt context * train-text-from-scratch: automatically allocate input tensors * train-text-from-scratch: automatically allocate compute memory * remove unused options and equalize train-text-from-scratch with finetune * initialize opt->loss_after with zero * add export-lora program * remove trailing whitespace * add export-lora build in Makefile * remove unused struct tensor_info from export-lora * add export-lora build dependency to llama because it depends on common, which depends on llama * update finetune README.md * cancel optimization when specified number of epochs is completed * improve handling of export-lora arguments print errors and warnings when files could not be read or created * Fix export-lora.cpp "not enough space in the context's memory pool" (#1) * Fix export-lora.cpp "not enough space in the context's memory pool" Without this patch, export-lora would sometimes error with "not enough space in the context's memory pool (needed 656784, available 656800)". * increase required context size by 5*GGML_MEM_ALIGN instead of plain 16 --------- Co-authored-by: xaedes <xaedes@gmail.com> * improve handling of not yet supported tensor types --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: meatbag-18a <145869052+meatbag-18a@users.noreply.github.com>
2023-09-28 18:40:11 +00:00
const struct llama_model & model, const char * path_lora, float scale, const char * path_base_model, int n_threads
check C++ code with -Wmissing-declarations (#3184)
2023-09-15 19:38:27 +00:00
) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
LLAMA_LOG_INFO("%s: applying lora adapter from '%s' - please wait ...\n", __func__, path_lora);
const int64_t t_start_lora_us = ggml_time_us();
lora : add support for non-llama models (#3333) * lora : add support for non-llama models ggml-ci * avoid leaking ggml_context on failure cleanup ggml-ci * lora : allow 1d tensors * lora : include embd and output layers in size calculation * fix style
2023-12-16 17:58:46 +00:00
llama_file fin(path_lora, "rb");
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
// verify magic and version
{
lora : add support for non-llama models (#3333) * lora : add support for non-llama models ggml-ci * avoid leaking ggml_context on failure cleanup ggml-ci * lora : allow 1d tensors * lora : include embd and output layers in size calculation * fix style
2023-12-16 17:58:46 +00:00
uint32_t magic = fin.read_u32();
if (magic != LLAMA_FILE_MAGIC_GGLA) {
LLAMA_LOG_ERROR("%s: bad file magic\n", __func__);
return 1;
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
lora : add support for non-llama models (#3333) * lora : add support for non-llama models ggml-ci * avoid leaking ggml_context on failure cleanup ggml-ci * lora : allow 1d tensors * lora : include embd and output layers in size calculation * fix style
2023-12-16 17:58:46 +00:00
uint32_t format_version = fin.read_u32();
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
if (format_version != 1) {
LLAMA_LOG_ERROR("%s: unsupported file version\n", __func__ );
return 1;
}
}
lora : add support for non-llama models (#3333) * lora : add support for non-llama models ggml-ci * avoid leaking ggml_context on failure cleanup ggml-ci * lora : allow 1d tensors * lora : include embd and output layers in size calculation * fix style
2023-12-16 17:58:46 +00:00
int32_t lora_r = fin.read_u32();
int32_t lora_alpha = fin.read_u32();
train : finetune LORA (#2632) * fix track_max_mem in forward_batch_wo_cache_flash_attn_train * remove unnecessary Adam(W) optimizer tensors. reduces optimizer memory overhead from 7*modelsize to 2*modelsize. additionally allows to optimize models with more than 2^31 parameters by replacing int with int64_t. bumps training checkpoint file version, but old checkpoints can still be read. new version with less tensors is saved. * add gradient clipping to AdamW * Fix reset of unused g->nodes and g->grads to NULL * implement gradient checkpointing for training reduces memory overhead from O(n_layer) to O(sqrt(n_layer)) as explained in readme of https://github.com/cybertronai/gradient-checkpointing * remove unused compute buffer 3 * add and use function ggml_build_backward_expand to avoid stack overflows with large maximum number of nodes GGML_API void ggml_build_backward_expand(struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, bool keep); * change AdamW decay parameter to work like the torch AdamW decay parameter It is now relative to Adam learning rate `alpha*sched`. Before that it was relative to `sched` only. `alpha` being the maximum learning rate and `sched` being a scaling parameter in [0..1] * change default AdamW weight decay parameter used in training to 0.1 as used in nanoGPT * change default AdamW weight decay parameter defined in ggml to 0.0, making Adam default instead of AdamW btw: the default weight decay parameter for torch.optim.AdamW is 0.01 * bug fixes for cross entropy loss ggml_cross_entropy_loss: sums where not correctly added in workload of each thread ggml_cross_entropy_loss_back: simplify backward process, reducing numerical issues guard usage of exp f16 lookup in cross entropy by #define GGML_CROSS_ENTROPY_EXP_FP16 cross entropy loss is only used once during training, but it is quite sensitive to numerical errors introduced by exp-f16-lookup. so exp-f16-lookup for cross entropy loss is disabled by default, trading better gradients for very slightly worse runtime performance. * fix test-grad0 for cross_entropy_loss the second argument to cross_entropy_loss must sum up to 1 for each row * fix test-grad0 for soft_max dont use only sum as aggregation, because sum of softmax is always 1 -> finite differences should not work instead use sum(log(soft_max()*(1-eps)+eps)); use eps to avoid log(0) * improve finite differences of test-grad0 by using double instead of float * change cross_entropy_loss to output average over all rows this helps keeping the loss and gradients in a sane range * improve gradient checkpointing sqrt(n_layers) is only the best checkpoint step when mem size of checkpoints and mem size of layers are equal. since layers require more memory than the single-tensor-checkpoint we use, the optimal values are compute different: ``` given: n, u, v objective: minimize(a*u+b*v) where a*b=n, a>0, b>0 b=n/a minimize(a*u+v*n/a) diff(a*u+v*n/a, a) = u - (v*n/a)/a diff(a*u+v*n/a, a) == 0 u - (v*n/a)/a == 0 u == v*n/(a*a) u*a*a = v*n a*a = v*n/u a = sqrt(n*v/u) ``` this change results in more checkpoints, requiring less layers to store between checkpoints, overall improving memory usage. * disable gradient checkpointing debug output * llama : fix rope usage in train-text-from-scratch after ChatGLM change * add more training parameters: --enable-restart N Only for Adam optimizer. Enable restarts of cos-decay --disable-restart N Only for Adam optimizer. Disable restarts of cos-decay --opt-past N Number of optimization iterations to track for delta convergence test. Disabled when zero. --opt-delta N Maximum delta for delta convergence test. Disabled when <= zero. --opt-max-no-improvement N Maximum number of optimization iterations with no improvement. Disabled when <= zero. --adam-epsf N AdamW epsilon for convergence test. Disabled when <= zero. --adam-min-alpha N Adam minimum learning rate alpha, usually 0.1 * alpha * replace memcpy with reshape operation so that the graph is not cut at the input this makes it possible to store other values into the input tensor and then simply recompute the graph without rebuilding it * remove unused function argument from get_example_targets_batch * measure and print total training time * add optimization callback to ggml_opt_resume_g this callback is called before each iteration with custom data and pointer to learning schedule parameter (only used in Adam(W)). can be used for dynamic learning schedule and setting input data for batches before each iteration * use optimization callback in training allows dynamic learning schedule and different batch data for each iteration without relying on low n_iter and high n_examples parameters reduces runtime by avoiding restart of optimization function and improves training convergence by providing a different batch for each iteration * add minimum number of tensor dimensions to apply weight decay (default 2) this allows to not apply weight decay to bias parameters * rename training parameter cos-decay-alpha to cos-decay-min and clarify that adam-min-alpha also applies to warmup * fix increase of model.train_samples and model.train_tokens now that each optimizer iteration gets its own batch we need to multiply by number of opt iterations * change sampling parameters for prediction after training to defaults of common.h and clarify what is context for prediction and what are generated tokens * tighten abs error bounds for cross_entropy_loss in test-grad0 * add conditional compilation of using F16 exp in flash attention uncomment `// #define GGML_FLASH_ATTN_EXP_FP16` to enable usage of f16 exp in flash attention * tighten abs error bounds for flash_attn in test-grad0 * tighten abs error bounds for sqrt in test-grad0 * remove out-commented vectorized code of opt_adam the vectorized code might be bit faster for low number of parameters, but it had a big memory usage overhead * ggml : update ggml_rms_norm_back with configurable eps * llama training : fix ggml_rms_norm_back calls to pass configurable eps * remove trailing whitespace * add train function using automatic gradient checkpointing backward pass and allocator * in train function replace add_inplace by regular add because using add_inplace seems to result in different gradients * don't use allocate hash_map on context because the context has no_alloc=True when using memory allocator resulting in NULL data pointers * correctly clone reshape and permute operations by also cloning tensor->nb values * fix variable name and add missing type cast * terminate recursive tensor cloning when reaching tensor without src tensors * correctly clone view tensors by setting data pointers without this the checkpointing would only work when being used together with memory allocator * fix variable names * swap arguments to commutative ops to be the same as in `forward_batch_wo_cache_flash_attn` * add input tensors as checkpoints so that recursive tensor cloning of gradient checkpointing terminates on input tensors * fix variable name and add missing boolean negation * make sure some tensors are not reallocated by inserting new temporary nodes depending on them: output and parameter gradient tensors need to be available at the end of the graph execution parameter gradient tensors also need to be available before the graph execution because they are set to zero before each optimizer iteration checkpoint tensors are allocated all together to reduce memory allocator fragmentation afterwards, in addition to the temporary nodes, we also need to reset the temporary leafs * fix ASSERT to work with zero layers * add training options whether to use allocator and/or unified training function * integrate unified training function which may use memory allocator the unified training function also supports arguments whether to use flash attention and/or gradient checkpointing * format name of cloned tensors with " (clone)" suffix * set names for tensors in unified train function for easier debugging * allocate graph on context using ggml_new_graph * remove handwritten training functions * remove unused training parameters "use_scratch" and "use_unified" * remove trailing whitespace * remove unused train params: mem_compute1_gb & mem_compute2_gb mem_compute_gb is used for compute when automatic memory allocator is not enabled, otherwise it can be very small to only hold the tensor definitions mem_compute0_gb is used for automatic memory allocator (as long as measurement of max required size is not implemented) * remove unused forward_batch function * add debug asserts in ggml_allocr_alloc to some common pitfalls when using this function directly * only use ggml_allocr_alloc when tensor has NULL data and is no view * fix test when to create temporary backward graph temporary backward graph is only necessary when using checkpointing * fix memory "leak" in optimizers each iteration a new cplan with new memory for work data was allocated. now cplan creation only happens at the start of optimization, with each iteration reusing the cplan and its work data. * reverse order of for loop in ggml_build_backward_expand to save memory when using gradient checkpointing and allocator with this loop order gradient checkpointing with allocator on 16 layer model saves 13% memory; 2 layer memory it saves 2% memory. the computation results are the same * add API functions to access llama model tensors * add stub example for finetuning, based on train-text-from-scratch * move and remove code * add API functions to access remaining model parameters: mult, head and rot * first draft for LORA finetune training * remove const model and layer arguments in API functions for accessing model tensors * bug fixes to make finetune compile automatic allocator does not work yet * add debug prints for training memory improvements * fix names of lora tensors * avoid stack overflow resulting from big ggml_cgraph replace stack allocation and ggml_build_forward by ggml_new_graph in combination with ggml_build_forward_expand * replace llama API functions to get model tensors by one function to get model tensor by name LLAMA_API struct ggml_tensor * llama_get_model_tensor(struct llama_model * model, const char * name); * remove unused call to not existing llama_get_layer_from_model * implement ggml_compute_forward_out_prod_q_f32 * remove trailing whitespace * add lora finetune support on quantized base model tensors * add ggml_add_cast API function this function works like ggml_add, but accepts a data type for the resulting tensor. only supported for quantized src0 input. * use ggml_add_cast in finetuning lora-applied weights will now have data type F32, which improves gradients when finetuning quantized base models * bug fix: actually use result type passed to ggml_add_cast * make sure base model tensors data cannot be used in viewable operations memory allocator would try to make lora application inplace on base model tensors. since those are memory mapped this will result in memory access violations * fix bug in ggml_out_prod which resulted in wrong n_dims of result tensors * avoid keeping in memory ALL of the gradients The problem here stems from ggml_graph_reset. This function is called in the optimization function, before each graph computation, to reset the gradients to zero. This required a unique memory slot for each gradient: allocating memory from a previosly freed memory location might lead to non-zero input gradients. During ggml_compute_backward the gradients are build stepwise by adding or substracting new values, starting from a OP_NONE tensor which needs to contain zero-values. This requires the graph reset. To avoid this I now remember in ggml_build_backward_expand the original OP_NONE gradient tensors in a hash table, which is passed to ggml_compute_backward. There instead of using add (or sub or similar) I test whether the existing gradient to be changed is a zero-valued-tensor by looking up its existence in the hash table. When it is such a zero-tensor it will not be modified, but replaced by the value to be added, otherwise the regular add (not inplace, allocator will take care of this) will be used. This way none of those zero-tensor values will be necessary in the final backward graph and more importantly they won't need a unique memory slot, just to make them zero. * remove trailing whitespace * remove debug prints and function to compute tensor data hash * improve optimization iteration prints * adjust maximal values to support finetuning 3B models * change default finetune params lora_r and lora_alpha to match the n_rank parameters of 4 * bug fix: make sure finetune input gradient is allocated at begin and kept until end * remove unnecessary src tensor from ggml_get_rows_back we don't need data of src[2] for computation, only to setup the correct output shape. remove dependency on src[2], so that allocator can work more freely. the computational graph is still completely determined, because the output shape is naturally included. this is similar to how ggml_reshape does it. * remove unnecessary src tensor from ggml_repeat & ggml_repeat_back we don't need data of src[1] for computation, only to setup the correct output shape. remove dependency on src[1], so that allocator can work more freely. the computational graph is still completely determined, because the output shape is naturally included * resolve todo allocator will only make it inplace when they are of the same type * mixing multiple LORA adapters is now possible pass more than one '--lora FNAME' argument to apply more than one LORA. use '--lora-scaled FNAME S' when you want to specify a user-defined scale for an adapter. * add option to save finetune output every N iterations * also save latest finetune output with ITERATION="LATEST" and print where files are saved saving with LATEST makes it easier to resume training from the latest checkpoint the string "LATEST" can be configured with command line option "--fn-latest STR" * update checkpoint train stats before saving via "--save-every" * add command line option `--rank-wo N` for rank of wo tensor * update finetune README * fix dump_non_result_info_yaml to output multiple lora adapters * bug fix: replace GGML_TYPE_SIZE[t] by ggml_type_size(t) * replace llama_n_mult by llama_n_ff * finetune bug fixes to compile with merged in code from master * remove prediction related code to reduce duplicated code with main use main instead * reduce large memory overhead in train-text-from-scratch all gradients had to be pinned so that graph_reset works correctly. this is no longer necessary with the changes to ggml_compute_backward introduced in this PR. * add comment explaining why finetune checkpoints are allocated in one block * make default value of float member a float literal * handle rms_norm and rope parameters the same as in train-text-from-scratch * remove unused code * remove vocab related code as it is unnecessary * add LLM_KV_TRAINING_TYPE to train-text-from-scratch checkpoints so that they can be differentiated from lora finetune checkpoints * add gguf constants and load/save functions from train-text-from-scratch * add load & save lora finetune checkpoints via gguf * add python script to convert old finetune checkpoint files to gguf * remove old checkpoint save & load code * remove code to print data checksums which was used to verify correctness of new gguf code * omit tokenization when training is disabled, only save llama lora adapter training can be disabled by passing '-n 0' to finetune * remove trailing whitespace * update README.md * implement ggml_compute_forward_repeat_f16 * avoid stack overflow of large cgraphs in test-grad0 * add ggml API functions ggml_unravel_index, ggml_get_i32_nd and its analogs for set and for f32 ggml_get_i32_1d, ggml_set_i32_1d, ggml_get_f32_1d, ggml_set_f32_1d now support non-contiguous tensors. in case of non-contiguous tensor, the 1d index is unraveled into a multi index using ggml_unravel_index to be passed to '_nd' function equivalent. this fixes a bug in test-grad0 which happens due to ggml_build_backward not building purely contiguous tensors anymore * increase test-grad0 context mem size to accommodate for bigger cgraph * add sanity check to ggml_compute_backward, asserting the correct shape of gradients * fix ggml_acc_or_set to return tensor of correct shape * remove unused 'inplace' argument from ggml_compute_backward function inplace operations to add gradients are no longer created by ggml_compute_backward use allocator to automatically make inplace operations * add missing argument 'int i0' to ggml_get_i32_nd & ggml_set_i32_nd header declarations * fix error message in ggml_allocr_alloc to display actual max_avail * fix check_gradient ggml_build_backward_expand was previously replaced by ggml_build_backward, but the assignment of forward graph to backward graph missing * use tensor->view_src instead of ggml_is_view and get_view_source * move gradient checkpointing code into ggml, new API function: // build gradient checkpointing backward graph gb for gf using provided checkpoints // gb_tmp will contain original backward graph with rewritten backward process nodes, // but without the second forward pass nodes. GGML_API void ggml_build_backward_gradient_checkpointing( struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, struct ggml_cgraph * gb_tmp, struct ggml_tensor * * checkpoints, int n_checkpoints); * replace custom data getters and setters by ggml functions * train-text-from-scratch can train (full finetune) gguf models just pass the gguf model via `--checkpoint-in FN`. after this, to continue training, pass the generated checkpoint instead of the original gguf model. tested with smaller models, bigger models may exceed available memory. use (LORA) finetune for those. * remove trailing whitespace * add option to save train-text-from-scratch output every N iterations * update README.md * fix warnings * fix warnings * remove finetune option to disable allocator the allocator should always be used. by making sure that it is always used it gets easier to implement automatic memory requirements computation * add tensor checkpoints only when gradient checkpointing is enabled * initialize opt ggml context if none was provided * add ggml-alloc API function 'ggml_allocr_max_size' to get max size of alloc GGML_API size_t ggml_allocr_max_size(struct ggml_allocr * alloc); * finetune: automatically allocate all memory and changes to command line options remove '--n_examples N' parameter, as it no longer makes sense to call optimization process multiple times in a loop. add '--only_write_lora' command line option: will skip tokenization and training, to only write a llama.cpp comptabile LORA adapter. remove memory buffer related command line options. improve iteration console output. * add finetune to Makefile * update README.md * print time per iteration and estimate remaining time * increase measured alloc size by tensor_alignment ggml_allocr_reset will reduce the given size by up to tensor_alignment-1 * fix README.md * add some more allocator debug prints * bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue * revert last commit "bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue" "alloc was freeing an externally allocated tensor, because it calculated the end of allocator memory as alloc->data + alloc->max_size instead of alloc->data + alloc->size." This is intentional to reduce the risk of freeing external tensors when measuring. Unless max_size is not properly calculated, I don't see why this is an issue. * remove unnecessary "0x" before "%p" output * move measurement memory segment to upper region of the address space * update README.md * fix printf format warnings * add missing gguf_free in load_checkpoint_lora_file * load default rms_norm and rope parameters from base model * add gradient accumulation specify number accumulation steps with '--grad-acc N'. this will simulate a bigger batch size of grad_acc*batch. * fix tracking of train_samples and train_tokens * build : fix compile warnings * ggml : fix L-BFGS linesearch loop * improve finetune time measurement fix printf warnings on system where int64_t is (long int). change time datatypes to double because values get big with long training times. exclude file saving from time measurement. converge faster to actual time per iteration by removing very small first duration before first iteration was performed. fix bug in output of total training time, the reported value was 1000 times to small. * specify default lora rank with '--lora-r N' '--lora-r N' will specify default rank for all tensors '--rank-wq N', etc. will override this default rank for specific tensor types. * fix gradient accumulation bug where the same batch was used for each microstep * fix gradient accumulation bug where the same batch was used for each microstep * support grouped-query-attention in ggml_flash_attn and ggml_flash_attn_back k and v can now be repeated in q along ne[2] in forward pass just use modulo to compute k and v indices, like ik2 = iq2 % nek2. in backard pass this won't work as easy, because multiple threads will compete to accumulate to the same k->grad[:,ik1,ik2,ik3] and v->grad[:,iv1,iv2,iv3]. so we change the parallelization over q rows to be over k rows. this ensures non-overlapping (ik2,ik3) across threads. in each thread we then iterate over the number of repetitions of k/v in q to compute iq2 as iq2 = ik2 + irep*nek2. since ne2 is not the same for q,k and v we also change how the gradients are concatenated into the result tensor. additionally the offsets of gradq, gradk and gradv in the result tensor are now memory aligned. we also simplify the compute_backward part of flash_attn to use ggml_reshape instead of switching over the number of dimensions. this needs a small change to ggml_reshape, removing the assertion of second argument to be contiguous. since only the shape (ne) of the second reshape argument is of relevance, its memory layout (nb) is irrelevant -> it can very well be non-contiguous. change test-grad0 to also test for repeated k/v in q. this changes the rng and now results in small gradient differences in softmax. these solely come from using f16 exp table lookup in forward softmax: when temporarily changing softmax to use actual exp function, the reported gradient differences go away. gradient differences coming solely from f16 table lookup are acceptable. added a note to explain this. * add llama API functions to get grouped-query-attention n_head parameter 'n_head_kv'. * fix finetune to support grouped-query-attention (using flash-attention) note: ggml changes to ggml_out_prod are necessary to support grouped-query-attention without flash-attention. * support broadcastable a in out_prod(a, b) and backward pass of broadcasting mul_mat(a, b) * test broadcasting mul_mat backward pass * decouple random number generator of each operation test when changing one test the rng of others tests is not influenced anymore * add comment briefly describing what ggml_repeat_back does * simplify broadcasting mul_mat backward using ggml_repeat_back * add cgraph evaluation order member and corresponding enum type this controls in which order ggml_build_forward visits source nodes. by default the nodes are visited left to right, i.e. src[0] first. in some cases it is beneficial for ggml-alloc to visit in a different order. two possible orders are supported: left-to-right (src[0] first) and right-to-left (src[0] last). * measure max compute size for each cgraph eval order and use best order this can bring huge memory savings: e.g. codellama-34b with n_ctx=64, n_batch=1 goes from 92927.8mb down to 4627.6 MB * remove unused command line options * add sample start patterns and options to force new or by default resume last shuffling * update shuffle rng state on reshuffle * exclude known zero values from computations in flash_attn_f32 & flash_attn_back_f32 * remove probably unnecessary exception type flags from stringstream * pass correct max number of tokens to llama_tokenize * account for possible leading whitespace that will be added by tokenizer e.g. '\t' will be tokenized by llama spm tokenizer to [29871, 12] * use unrolled vec_mad in out_prod y is vec_mad result vec. x is vec_mad input vec. v is vec_mad input scalar. ggml_vec_mad_f32_unroll will internally loop over x and v with same y. GGML_VEC_MAD_UNROLL is by default defined to 32. This value is empirical optimized using performance test runs of out-prod in openllama-3b finetune with 256 context length and batch size 1. It gives 23% performance boost for out_prod. Full measurements of out-prod runtime in ms: unroll_xv unroll_yv 1 67014.643 87826.469 2 77117.552 89077.656 4 72091.311 109121.657 8 61077.543 88678.334 16 56914.67 79514.947 24 59024.595 84350.254 28 55952.446 83368.73 32 51476.658 85177.745 36 55973.792 84659.92 40 55139.616 93844.738 48 60736.392 93330.267 64 99856.878 116994.99 Second column is when unrollying yv instead of xv * set lora_alpha to value of lora_r if it is not set via command line otherwise only changing lora_r will change scaling of lora adapter used in prediction * reshuffle original sample order instead of the previous shuffled order otherwise resumed reshuffle will not result in same sample order * block tiling for out-prod inspired by mul-mat block sizes are empirically optimized roughly doubles the flops of out-prod * exclude some more known zero values from computations in flash_attn_f32 & flash_attn_back_f32 * add static keywords * remove outcommented old code * update train-text-from-scratch with tokenization, sample selection and shuffling from finetune * remove lbfgs related train parameters * move common train functions into common/train.[h|cpp] * move train state into struct train_state * move train data saving code into callback to unify code of opt_callback train_params are still different in finetune and train-text-from-scratch, so it can't yet be moved to train.h|cpp * move common train params into common/train * move common opt_callback into common/train * fix consume_common_train_arg * save and load head_count_kv in lora checkpoints * increase train_samples by used_samples instead of number of batches on batch can contain more than one sample when option "fill_with_next_samples" is used * fix usage of llama_tokenize * remove static from process_escape since we need it exposed in header * fix code formating of long function declarations * fix condition in load_train_state_gguf * use die("msg") instead of replace GGML_ASSERT(!"msg") or throw std::runtime_error("msg") * fix saving and loading of training type * remove terminating '\0' from tokenization (llama_tokenize is now passed the string length instead of relying on terminating '\0') * fix compile warnings * fix compile warnings * use new/delete for train_state instead of malloc/free using malloc may result in seg faults when trying to assign string fields * assert that sample_count > 0, avoiding division by zero * fix frand to return value in interval [0,1) * add train option "--sample-random-offsets" Use samples beginning at random offsets. The offset is only applied to the first sample in each batch context window. Together with "--fill-with-next-samples" this may help for training endless text generation. For example given a dataset containing samples "abcd", "ABCD", "0123". With context size of 8 and options "--fill-with-next-samples", "--no-separate-with-eos", "--no-separate-with-bos", the context windows of batches could only be filled with "abcdABCD", "ABCDabcd", "0123abcd", etc. With "--sample-random-offsets" it can also be filled with "23abcdAB", "bcd0123A", etc. * deduplicate code into function * remove n_rot hparam, as it must always be hparam.n_embd_head() * align code * assert correct base model tensor shapes * move some params from lora hparams into model hparams and load model params from gguf this equalizes the model definition in finetune and text-from-scratch and removes the need for additional llama api functions to get model parameters * remove now unnecessary llama API functions to get model params that where added by this PR * train-text-from-scratch: automatically allocate model tensors, remove option '--mem-model N' * train-text-from-scratch: automatically allocate opt context * train-text-from-scratch: automatically allocate input tensors * train-text-from-scratch: automatically allocate compute memory * remove unused options and equalize train-text-from-scratch with finetune * initialize opt->loss_after with zero * add export-lora program * remove trailing whitespace * add export-lora build in Makefile * remove unused struct tensor_info from export-lora * add export-lora build dependency to llama because it depends on common, which depends on llama * update finetune README.md * cancel optimization when specified number of epochs is completed * improve handling of export-lora arguments print errors and warnings when files could not be read or created * Fix export-lora.cpp "not enough space in the context's memory pool" (#1) * Fix export-lora.cpp "not enough space in the context's memory pool" Without this patch, export-lora would sometimes error with "not enough space in the context's memory pool (needed 656784, available 656800)". * increase required context size by 5*GGML_MEM_ALIGN instead of plain 16 --------- Co-authored-by: xaedes <xaedes@gmail.com> * improve handling of not yet supported tensor types --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: meatbag-18a <145869052+meatbag-18a@users.noreply.github.com>
2023-09-28 18:40:11 +00:00
float scaling = scale * (float)lora_alpha / (float)lora_r;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
LLAMA_LOG_INFO("%s: r = %d, alpha = %d, scaling = %.2f\n", __func__, lora_r, lora_alpha, scaling);
// load base model
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
std::unique_ptr<llama_model_loader> ml;
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
if (path_base_model) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
LLAMA_LOG_INFO("%s: loading base model from '%s'\n", __func__, path_base_model);
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
ml.reset(new llama_model_loader(path_base_model, /*use_mmap*/ true, /*check_tensors*/ false, /*kv_overrides*/ nullptr));
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
ml->init_mappings(/*prefetch*/ false); // no prefetching
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
struct tensor_meta {
std::string name;
ggml_type type;
int32_t ne[2];
size_t offset;
};
std::map<std::string, tensor_meta> tensor_meta_map;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// load all tensor meta
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
while (true) {
lora : add support for non-llama models (#3333) * lora : add support for non-llama models ggml-ci * avoid leaking ggml_context on failure cleanup ggml-ci * lora : allow 1d tensors * lora : include embd and output layers in size calculation * fix style
2023-12-16 17:58:46 +00:00
if (fin.tell() == fin.size) {
// eof
break;
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
int32_t n_dims;
lora : add support for non-llama models (#3333) * lora : add support for non-llama models ggml-ci * avoid leaking ggml_context on failure cleanup ggml-ci * lora : allow 1d tensors * lora : include embd and output layers in size calculation * fix style
2023-12-16 17:58:46 +00:00
int32_t name_len;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
int32_t ftype;
lora : add support for non-llama models (#3333) * lora : add support for non-llama models ggml-ci * avoid leaking ggml_context on failure cleanup ggml-ci * lora : allow 1d tensors * lora : include embd and output layers in size calculation * fix style
2023-12-16 17:58:46 +00:00
fin.read_raw(&n_dims, sizeof(n_dims));
fin.read_raw(&name_len, sizeof(name_len));
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
fin.read_raw(&ftype, sizeof(ftype));
lora : add support for non-llama models (#3333) * lora : add support for non-llama models ggml-ci * avoid leaking ggml_context on failure cleanup ggml-ci * lora : allow 1d tensors * lora : include embd and output layers in size calculation * fix style
2023-12-16 17:58:46 +00:00
if (n_dims != 1 && n_dims != 2) {
LLAMA_LOG_ERROR("%s: unsupported tensor dimension %d\n", __func__, n_dims);
return 1;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
int32_t ne[2] = { 1, 1 };
for (int i = 0; i < n_dims; ++i) {
lora : add support for non-llama models (#3333) * lora : add support for non-llama models ggml-ci * avoid leaking ggml_context on failure cleanup ggml-ci * lora : allow 1d tensors * lora : include embd and output layers in size calculation * fix style
2023-12-16 17:58:46 +00:00
fin.read_raw(&ne[i], sizeof(ne[i]));
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
std::string name;
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
GGML_ASSERT(name_len < GGML_MAX_NAME);
char buf[GGML_MAX_NAME];
lora : add support for non-llama models (#3333) * lora : add support for non-llama models ggml-ci * avoid leaking ggml_context on failure cleanup ggml-ci * lora : allow 1d tensors * lora : include embd and output layers in size calculation * fix style
2023-12-16 17:58:46 +00:00
fin.read_raw(buf, name_len);
name = std::string(buf, name_len);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// check for lora suffix
std::string lora_suffix;
if (name.length() > 6) {
lora_suffix = name.substr(name.length() - 6);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
if (lora_suffix != ".loraA" && lora_suffix != ".loraB") {
LLAMA_LOG_ERROR("%s: error: '%s' is not a lora tensor\n", __func__, name.c_str());
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
return 1;
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// tensor type
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
ggml_type wtype;
switch (ftype) {
case 0: wtype = GGML_TYPE_F32; break;
case 1: wtype = GGML_TYPE_F16; break;
default:
{
LLAMA_LOG_ERROR("%s: invalid tensor data type '%d'\n",
__func__, ftype);
llama : minor fixed return int value (#5529)
2024-02-16 11:45:48 +00:00
return 1;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// data offset
lora : add support for non-llama models (#3333) * lora : add support for non-llama models ggml-ci * avoid leaking ggml_context on failure cleanup ggml-ci * lora : allow 1d tensors * lora : include embd and output layers in size calculation * fix style
2023-12-16 17:58:46 +00:00
size_t offset = fin.tell();
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
offset = (offset + 31) & -32;
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// skip tensor data
fin.seek(offset + ggml_row_size(wtype, ne[0]) * ne[1], SEEK_SET);
tensor_meta_map.emplace(name, tensor_meta{ name, wtype, { ne[0], ne[1] }, offset });
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
bool warned = false;
int n_tensors = 0;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// apply
ggml_backend_t backend_cpu = ggml_backend_cpu_init();
if (backend_cpu == nullptr) {
LLAMA_LOG_ERROR("%s: error: failed to initialize cpu backend\n", __func__);
return 1;
}
ggml_backend_cpu_set_n_threads(backend_cpu, n_threads);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
std::vector<no_init<uint8_t>> read_buf;
for (const auto & it : model.tensors_by_name) {
const std::string & base_name = it.first;
ggml_tensor * model_t = it.second;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
if (tensor_meta_map.find(base_name + ".loraA") == tensor_meta_map.end() ||
tensor_meta_map.find(base_name + ".loraB") == tensor_meta_map.end()) {
continue;
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
tensor_meta & metaA = tensor_meta_map.at(base_name + ".loraA");
tensor_meta & metaB = tensor_meta_map.at(base_name + ".loraB");
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ggml_init_params lora_init_params = {
/* .mem_size */ ggml_tensor_overhead()*128 + ggml_graph_overhead(),
/* .mem_buffer */ nullptr,
/* .no_alloc */ true,
};
ggml_context * lora_ctx = ggml_init(lora_init_params);
if (lora_ctx == nullptr) {
LLAMA_LOG_ERROR("%s: error: failed to initialize lora context\n", __func__);
ggml_backend_free(backend_cpu);
return 1;
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// create tensors
ggml_tensor * loraA = ggml_new_tensor_2d(lora_ctx, metaA.type, metaA.ne[0], metaA.ne[1]);
ggml_tensor * loraB = ggml_new_tensor_2d(lora_ctx, metaB.type, metaB.ne[0], metaB.ne[1]);
ggml_set_name(loraA, metaA.name.c_str());
ggml_set_name(loraB, metaB.name.c_str());
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ggml_tensor * base_t;
if (ml) {
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
if (!ml->get_tensor_meta(base_name.c_str())) {
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
LLAMA_LOG_ERROR("%s: error: tensor '%s' not found in base model\n", __func__, base_name.c_str());
return 1;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
base_t = ggml_dup_tensor(lora_ctx, ml->get_tensor_meta(base_name.c_str()));
} else {
base_t = ggml_dup_tensor(lora_ctx, model_t);
}
ggml_set_name(base_t, base_name.c_str());
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// allocate in backend buffer
ggml_backend_buffer_t lora_buf = ggml_backend_alloc_ctx_tensors_from_buft(lora_ctx, ggml_backend_cpu_buffer_type());
if (lora_buf == nullptr) {
LLAMA_LOG_ERROR("%s: error: failed to allocate lora tensors\n", __func__);
return 1;
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// load tensor data
auto load_tensor = [&read_buf, &fin](const tensor_meta & tensor_meta, ggml_tensor * tensor) {
read_buf.resize(ggml_nbytes(tensor));
fin.seek(tensor_meta.offset, SEEK_SET);
fin.read_raw(read_buf.data(), ggml_nbytes(tensor));
ggml_backend_tensor_set(tensor, read_buf.data(), 0, read_buf.size());
};
load_tensor(metaA, loraA);
load_tensor(metaB, loraB);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// load base model tensor data
if (ml) {
ml->load_data_for(base_t);
} else {
ggml_backend_tensor_copy(model_t, base_t);
}
if (ggml_is_quantized(base_t->type) && !warned) {
LLAMA_LOG_WARN("%s: warning: using a lora adapter with a quantized model may result in poor quality, "
"use a f16 or f32 base model with --lora-base\n", __func__);
warned = true;
}
if (base_t->ne[0] != loraA->ne[1] || base_t->ne[1] != loraB->ne[1]) {
LLAMA_LOG_ERROR("%s: incompatible tensor dimensions (%" PRId64 " and %" PRId64 ");"
" are you sure that this adapter is for this model?\n", __func__, base_t->ne[0], loraA->ne[1]);
ggml_free(lora_ctx);
ggml_backend_buffer_free(lora_buf);
ggml_backend_free(backend_cpu);
return 1;
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
auto build_lora_graph = [&]() {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
// w = w + BA*s
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ggml_tensor * BA = ggml_mul_mat(lora_ctx, loraA, loraB);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
ggml_set_name(BA, "BA");
if (scaling != 1.0f) {
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
BA = ggml_scale(lora_ctx, BA, scaling);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
ggml_set_name(BA, "BA_scaled");
}
ggml_tensor * r;
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
r = ggml_add_inplace(lora_ctx, base_t, BA);
ggml_set_name(r, "r_add");
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
if (base_t->type != model_t->type) {
// convert the result to the model type
r = ggml_cast(lora_ctx, r, model_t->type);
ggml_set_name(r, "r_cast");
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
return r;
};
ggml_cgraph * gf = ggml_new_graph(lora_ctx);
ggml_tensor * r = build_lora_graph();
ggml_build_forward_expand(gf, r);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ggml_backend_buffer_t graph_buf = ggml_backend_alloc_ctx_tensors_from_buft(lora_ctx, ggml_backend_cpu_buffer_type());
if (graph_buf == nullptr) {
LLAMA_LOG_ERROR("%s: error: failed to allocate graph tensors\n", __func__);
ggml_free(lora_ctx);
ggml_backend_buffer_free(lora_buf);
ggml_backend_free(backend_cpu);
return 1;
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ggml_backend_graph_compute(backend_cpu, gf);
lora : add support for non-llama models (#3333) * lora : add support for non-llama models ggml-ci * avoid leaking ggml_context on failure cleanup ggml-ci * lora : allow 1d tensors * lora : include embd and output layers in size calculation * fix style
2023-12-16 17:58:46 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ggml_backend_tensor_set(model_t, r->data, 0, ggml_nbytes(r));
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
#if 0
// TODO: use scheduler with fallback to CPU for less copies between CPU and GPU
//ggml_backend_sched_t sched = ggml_backend_sched_new(backends.data(), backends.size(), GGML_DEFAULT_GRAPH_SIZE);
// sched compute
ggml_build_forward_expand(gf, build_graph());
ggml_backend_sched_init_measure(sched, gf);
// create the graph again, since the previous one was destroyed by the measure
ggml_graph_clear(gf);
ggml_build_forward_expand(gf, build_graph());
ggml_backend_sched_graph_compute(sched, gf);
ggml_backend_sched_free(sched);
#endif
ggml_backend_buffer_free(lora_buf);
ggml_backend_buffer_free(graph_buf);
ggml_free(lora_ctx);
n_tensors++;
if (n_tensors % 4 == 0) {
LLAMA_LOG_INFO(".");
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ggml_backend_free(backend_cpu);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
const int64_t t_lora_us = ggml_time_us() - t_start_lora_us;
LLAMA_LOG_INFO(" done (%.2f ms)\n", t_lora_us / 1000.0);
return 0;
}
//
// interface implementation
//
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
struct llama_model_params llama_model_default_params() {
struct llama_model_params result = {
build : on Mac OS enable Metal by default (#2901) * build : on Mac OS enable Metal by default * make : try to fix build on Linux * make : move targets back to the top * make : fix target clean * llama : enable GPU inference by default with Metal * llama : fix vocab_only logic when GPU is enabled * common : better `n_gpu_layers` assignment * readme : update Metal instructions * make : fix merge conflict remnants * gitignore : metal
2023-09-04 19:26:24 +00:00
/*.n_gpu_layers =*/ 0,
code : normalize enum names (#5697) * coda : normalize enum names ggml-ci * code : cont * code : cont
2024-02-25 10:09:09 +00:00
/*.split_mode =*/ LLAMA_SPLIT_MODE_LAYER,
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
/*.main_gpu =*/ 0,
/*.tensor_split =*/ nullptr,
/*.progress_callback =*/ nullptr,
/*.progress_callback_user_data =*/ nullptr,
llama : allow overriding GGUF metadata when loading model (#4092) * feat: Allow overriding GGUF metadata when loading model * Fix the one time GCC is stricter than clang about something * Step1 * Refactor... basically everything! * Nuke obsolete GetArrayLen struct * simplify std::string specialization * Various cleanups Add informational output when overrides are applied Warn user when an override with the wrong type is specified * Fix broken logic for parsing bool KV overrides Fix issue where overrides didn't apply when key missing in GGUF metadata Resolve merge changes * llama : rearrange model params * Update new GET_KEY call Add note that metadata KV overrides aren't reflected in initial metadata KV info dump --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-05 17:19:18 +00:00
/*.kv_overrides =*/ nullptr,
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
/*.vocab_only =*/ false,
/*.use_mmap =*/ true,
/*.use_mlock =*/ false,
add basic tensor data validation function (#6884) * add basic tensor data validation function * add --check-tensors command line argument tensor validation is disabled by default and can be enabled by adding `--check-tensors` to the command line arguments. quantize always validates tensors.
2024-04-26 16:39:58 +00:00
/*.check_tensors =*/ false,
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
};
build : on Mac OS enable Metal by default (#2901) * build : on Mac OS enable Metal by default * make : try to fix build on Linux * make : move targets back to the top * make : fix target clean * llama : enable GPU inference by default with Metal * llama : fix vocab_only logic when GPU is enabled * common : better `n_gpu_layers` assignment * readme : update Metal instructions * make : fix merge conflict remnants * gitignore : metal
2023-09-04 19:26:24 +00:00
#ifdef GGML_USE_METAL
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// note: we usually have plenty of VRAM, so by default offload all layers to the GPU
result.n_gpu_layers = 999;
build : on Mac OS enable Metal by default (#2901) * build : on Mac OS enable Metal by default * make : try to fix build on Linux * make : move targets back to the top * make : fix target clean * llama : enable GPU inference by default with Metal * llama : fix vocab_only logic when GPU is enabled * common : better `n_gpu_layers` assignment * readme : update Metal instructions * make : fix merge conflict remnants * gitignore : metal
2023-09-04 19:26:24 +00:00
#endif
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
return result;
}
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
struct llama_context_params llama_context_default_params() {
struct llama_context_params result = {
/*.seed =*/ LLAMA_DEFAULT_SEED,
/*.n_ctx =*/ 512,
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
/*.n_batch =*/ 2048,
/*.n_ubatch =*/ 512,
llama : more consistent names of count variables (#5994) * llama : more consistent names of count variables ggml-ci * llama : n_parallel -> n_seq_max * common : fix param name * examples : fix param name
2024-03-11 15:49:47 +00:00
/*.n_seq_max =*/ 1,
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
/*.n_threads =*/ GGML_DEFAULT_N_THREADS, // TODO: better default
/*.n_threads_batch =*/ GGML_DEFAULT_N_THREADS,
code : normalize enum names (#5697) * coda : normalize enum names ggml-ci * code : cont * code : cont
2024-02-25 10:09:09 +00:00
/*.rope_scaling_type =*/ LLAMA_ROPE_SCALING_TYPE_UNSPECIFIED,
llama : allow for user specified embedding pooling type (#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-03 10:40:27 +00:00
/*.pooling_type =*/ LLAMA_POOLING_TYPE_UNSPECIFIED,
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
/*.rope_freq_base =*/ 0.0f,
/*.rope_freq_scale =*/ 0.0f,
llama : change yarn_ext_factor placeholder to -1 (#3922)
2023-11-03 06:31:58 +00:00
/*.yarn_ext_factor =*/ -1.0f,
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
/*.yarn_attn_factor =*/ 1.0f,
/*.yarn_beta_fast =*/ 32.0f,
/*.yarn_beta_slow =*/ 1.0f,
llama : fix llama_context_default_params after #2268 (#3893)
2023-11-01 23:29:14 +00:00
/*.yarn_orig_ctx =*/ 0,
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
/*.defrag_thold =*/ -1.0f,
backend : add eval callback (#4935) * backend : add eval callback ggml-ci * backend : group nodes in a single compute when user don't need them * backend : clean-up the implementation ggml-ci * simple : do not perform tensor data copy if not needed * simple : fix * simple : no need for ggml_is_contiguous + fix bool parse * llama : fix callback placement in llama_context_params * backend : avoid double-ask callback calls * simple : restore examples, imatrix will serve as a demo
2024-01-17 16:39:41 +00:00
/*.cb_eval =*/ nullptr,
/*.cb_eval_user_data =*/ nullptr,
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
/*.type_k =*/ GGML_TYPE_F16,
/*.type_v =*/ GGML_TYPE_F16,
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
/*.logits_all =*/ false,
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
/*.embeddings =*/ false,
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
/*.offload_kqv =*/ true,
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
/*.flash_attn =*/ false,
llama : add abort_callback to interrupt computation (#5409) * using abort_callback from ggml to stop llama computation * format fix * a brief explaining comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-02 19:52:25 +00:00
/*.abort_callback =*/ nullptr,
/*.abort_callback_data =*/ nullptr,
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
};
return result;
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
struct llama_model_quantize_params llama_model_quantize_default_params() {
struct llama_model_quantize_params result = {
/*.nthread =*/ 0,
/*.ftype =*/ LLAMA_FTYPE_MOSTLY_Q5_1,
quantize: options for output and token embedding tensors qtype (#6239) * quantize: be able to specify the output tensor type * quantize: be able to specify the token embedding tensor type --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-03-22 18:47:14 +00:00
/*.output_tensor_type =*/ GGML_TYPE_COUNT,
/*.token_embedding_type =*/ GGML_TYPE_COUNT,
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
/*.allow_requantize =*/ false,
/*.quantize_output_tensor =*/ true,
Allow quantize to only copy tensors, some other improvements (#2931) * Allow quantize tool to only copy tensors to allow repackaging models. * Slightly better logic when requantizing. * Change help message to go to `stdout`.
2023-09-01 14:02:48 +00:00
/*.only_copy =*/ false,
ggml : quantization refactoring (#3833) * ggml : factor all quantization code in ggml-quants ggml-ci * ggml-quants : fix Zig and Swift builds + quantize tool ggml-ci * quantize : --pure option for disabling k-quant mixtures --------- Co-authored-by: cebtenzzre <cebtenzzre@gmail.com>
2023-10-29 16:32:28 +00:00
/*.pure =*/ false,
quantize : add '--keep-split' to quantize model into shards (#6688) * Implement '--keep-split' to quantize model into several shards * Add test script * Update examples/quantize/quantize.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Split model correctly even if tensor id is out-of-order * Update llama_model_quantize_params * Fix preci failures --------- Co-authored-by: z5269887 <z5269887@unsw.edu.au> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-25 10:29:35 +00:00
/*.keep_split =*/ false,
2-bit quantizations (#4897) * imatrix: load * imatrix: WIP * imatrix: Add Q2_K quantization * imatrix: also guard against Q2_K_S quantization without importance matrix * imatrix: guard even more against low-bit quantization misuse --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
2024-01-14 07:45:56 +00:00
/*.imatrix =*/ nullptr,
quantize : be able to override metadata by key (#6321) * quantize: be able to override metadata by key * minor : spacing --------- Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 12:09:30 +00:00
/*.kv_overrides =*/ nullptr,
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
};
return result;
}
llama : remove LLAMA_MAX_DEVICES and LLAMA_SUPPORTS_GPU_OFFLOAD (#5240) * llama : remove LLAMA_MAX_DEVICES from llama.h ggml-ci * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * server : remove LLAMA_MAX_DEVICES ggml-ci * llama : remove LLAMA_SUPPORTS_GPU_OFFLOAD ggml-ci * train : remove LLAMA_SUPPORTS_GPU_OFFLOAD * readme : add deprecation notice * readme : change deprecation notice to "remove" and fix url * llama : remove gpu includes from llama.h ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-01-31 15:30:17 +00:00
size_t llama_max_devices(void) {
#if defined(GGML_USE_METAL)
return 1;
cuda : rename build flag to LLAMA_CUDA (#6299)
2024-03-26 00:16:01 +00:00
#elif defined(GGML_USE_CUDA)
llama : remove LLAMA_MAX_DEVICES and LLAMA_SUPPORTS_GPU_OFFLOAD (#5240) * llama : remove LLAMA_MAX_DEVICES from llama.h ggml-ci * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * server : remove LLAMA_MAX_DEVICES ggml-ci * llama : remove LLAMA_SUPPORTS_GPU_OFFLOAD ggml-ci * train : remove LLAMA_SUPPORTS_GPU_OFFLOAD * readme : add deprecation notice * readme : change deprecation notice to "remove" and fix url * llama : remove gpu includes from llama.h ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-01-31 15:30:17 +00:00
return GGML_CUDA_MAX_DEVICES;
#elif defined(GGML_USE_SYCL)
return GGML_SYCL_MAX_DEVICES;
Basic Vulkan Multi-GPU implementation (#5321) * Initial Vulkan multi-gpu implementation Move most global variables into backend context * Add names to backend device functions * Add further missing cleanup code * Reduce code duplication in tensor split layer assignment * generalize LLAMA_SPLIT_LAYER for all backends, do not expose device count and memory in llama.h * Only do device info print in the beginning and initialize one backend for cpu assist Add missing cleanup code * Rework backend memory management to make sure devices and buffers get properly allocated and freed * Rename cpu assist free function --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-07 06:54:50 +00:00
#elif defined(GGML_USE_VULKAN)
return GGML_VK_MAX_DEVICES;
llama : remove LLAMA_MAX_DEVICES and LLAMA_SUPPORTS_GPU_OFFLOAD (#5240) * llama : remove LLAMA_MAX_DEVICES from llama.h ggml-ci * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * server : remove LLAMA_MAX_DEVICES ggml-ci * llama : remove LLAMA_SUPPORTS_GPU_OFFLOAD ggml-ci * train : remove LLAMA_SUPPORTS_GPU_OFFLOAD * readme : add deprecation notice * readme : change deprecation notice to "remove" and fix url * llama : remove gpu includes from llama.h ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-01-31 15:30:17 +00:00
#else
return 1;
#endif
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llama : remove LLAMA_MAX_DEVICES and LLAMA_SUPPORTS_GPU_OFFLOAD (#5240) * llama : remove LLAMA_MAX_DEVICES from llama.h ggml-ci * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * server : remove LLAMA_MAX_DEVICES ggml-ci * llama : remove LLAMA_SUPPORTS_GPU_OFFLOAD ggml-ci * train : remove LLAMA_SUPPORTS_GPU_OFFLOAD * readme : add deprecation notice * readme : change deprecation notice to "remove" and fix url * llama : remove gpu includes from llama.h ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-01-31 15:30:17 +00:00
bool llama_supports_mmap(void) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
return llama_mmap::SUPPORTED;
}
llama : remove LLAMA_MAX_DEVICES and LLAMA_SUPPORTS_GPU_OFFLOAD (#5240) * llama : remove LLAMA_MAX_DEVICES from llama.h ggml-ci * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * server : remove LLAMA_MAX_DEVICES ggml-ci * llama : remove LLAMA_SUPPORTS_GPU_OFFLOAD ggml-ci * train : remove LLAMA_SUPPORTS_GPU_OFFLOAD * readme : add deprecation notice * readme : change deprecation notice to "remove" and fix url * llama : remove gpu includes from llama.h ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-01-31 15:30:17 +00:00
bool llama_supports_mlock(void) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
return llama_mlock::SUPPORTED;
}
llama : remove LLAMA_MAX_DEVICES and LLAMA_SUPPORTS_GPU_OFFLOAD (#5240) * llama : remove LLAMA_MAX_DEVICES from llama.h ggml-ci * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * server : remove LLAMA_MAX_DEVICES ggml-ci * llama : remove LLAMA_SUPPORTS_GPU_OFFLOAD ggml-ci * train : remove LLAMA_SUPPORTS_GPU_OFFLOAD * readme : add deprecation notice * readme : change deprecation notice to "remove" and fix url * llama : remove gpu includes from llama.h ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-01-31 15:30:17 +00:00
bool llama_supports_gpu_offload(void) {
cuda : rename build flag to LLAMA_CUDA (#6299)
2024-03-26 00:16:01 +00:00
#if defined(GGML_USE_CUDA) || defined(GGML_USE_CLBLAST) || defined(GGML_USE_METAL) || defined(GGML_USE_VULKAN) || \
defined(GGML_USE_SYCL) || defined(GGML_USE_KOMPUTE)
llama : remove LLAMA_MAX_DEVICES and LLAMA_SUPPORTS_GPU_OFFLOAD (#5240) * llama : remove LLAMA_MAX_DEVICES from llama.h ggml-ci * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * server : remove LLAMA_MAX_DEVICES ggml-ci * llama : remove LLAMA_SUPPORTS_GPU_OFFLOAD ggml-ci * train : remove LLAMA_SUPPORTS_GPU_OFFLOAD * readme : add deprecation notice * readme : change deprecation notice to "remove" and fix url * llama : remove gpu includes from llama.h ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-01-31 15:30:17 +00:00
// Defined when llama.cpp is compiled with support for offloading model layers to GPU.
return true;
#else
return false;
#endif
}
ggml : add numa options (#5377) * Added numa options to allow finer grained control as well as plumbing for a new mirror mode that will require numa.h * Reverted Makefile * Fixed include * Removed sched.h from ggml.h, moved ggml_get_numa_affinity into ggml.c, removed trailing whitespace and fixed up a few inconsistent variables * removed trailing whitespace * Added numa options to allow finer grained control as well as plumbing for a new mirror mode that will require numa.h * Reverting Makefile * Fixed a number of issues with the move from BOOL to ggml_numa_strategies. Added a note about mirror mode note being implemented yet * Removing MIRROR_MODE code for this PR * Removing last bit of MIRROR_MODE code for this PR * Removing unneeded branch in server.cpp example and moving get_numa_affinity and making it static * Fixed lingering init_llama_backend() bool calls in tests and examples * Remote enum llama_numa_strategies * Revert bad merge with dynatemp flags * add missing enum ggml_numa_strategies declaration and revert sync problem with master * add missing enum ggml_numa_strategies declaration * fixed ggml_init_numa variable * Update ggml.h Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Update READMEs with info about numa flags, change INTERLEAVE strategy name to DISTRIBUTE everywhere, implement the improved distribution strategy from @rankaiyx, fix a spelling mistake and un-merge some bad merges * split numa init out from llama_backend_init and created llama_numa_init. Updated all code paths and samples * Fix up some boolean vs enum comparisons * Added #ifdefs for non-Linux OS that don't have cpu_set_t datatype * Update ggml.h Align enum values Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml.c Remove whitespace Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml.c align paremeters Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update examples/server/server.cpp remove whitespace and align brace Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update common/common.cpp Remove whitespace and align brace Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * unified ggml_numa_strategy enum and fixed text alignment in server.cpp example * Update ggml.c simplified return for platforms without NUMA support Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * removed redundant else from cli argument processing of --numa * whitespace --------- Co-authored-by: root <root@nenya.lothlorien.ca> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Jared Van Bortel <jared@nomic.ai>
2024-02-16 09:31:07 +00:00
void llama_backend_init(void) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
ggml_time_init();
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
// needed to initialize f16 tables
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
{
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
struct ggml_init_params params = { 0, NULL, false };
struct ggml_context * ctx = ggml_init(params);
ggml_free(ctx);
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#ifdef GGML_USE_MPI
ggml_mpi_backend_init();
#endif
}
llama : support input embeddings directly (#1910) * add interface for float input * fixed inpL shape and type * add examples of input floats * add test example for embd input * fixed sampling * add free for context * fixed add end condition for generating * add examples for llava.py * add READMD for llava.py * add READMD for llava.py * add example of PandaGPT * refactor the interface and fixed the styles * add cmake build for embd-input * add cmake build for embd-input * Add MiniGPT-4 example * change the order of the args of llama_eval_internal * fix ci error
2023-06-28 15:53:37 +00:00
ggml : add numa options (#5377) * Added numa options to allow finer grained control as well as plumbing for a new mirror mode that will require numa.h * Reverted Makefile * Fixed include * Removed sched.h from ggml.h, moved ggml_get_numa_affinity into ggml.c, removed trailing whitespace and fixed up a few inconsistent variables * removed trailing whitespace * Added numa options to allow finer grained control as well as plumbing for a new mirror mode that will require numa.h * Reverting Makefile * Fixed a number of issues with the move from BOOL to ggml_numa_strategies. Added a note about mirror mode note being implemented yet * Removing MIRROR_MODE code for this PR * Removing last bit of MIRROR_MODE code for this PR * Removing unneeded branch in server.cpp example and moving get_numa_affinity and making it static * Fixed lingering init_llama_backend() bool calls in tests and examples * Remote enum llama_numa_strategies * Revert bad merge with dynatemp flags * add missing enum ggml_numa_strategies declaration and revert sync problem with master * add missing enum ggml_numa_strategies declaration * fixed ggml_init_numa variable * Update ggml.h Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Update READMEs with info about numa flags, change INTERLEAVE strategy name to DISTRIBUTE everywhere, implement the improved distribution strategy from @rankaiyx, fix a spelling mistake and un-merge some bad merges * split numa init out from llama_backend_init and created llama_numa_init. Updated all code paths and samples * Fix up some boolean vs enum comparisons * Added #ifdefs for non-Linux OS that don't have cpu_set_t datatype * Update ggml.h Align enum values Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml.c Remove whitespace Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update ggml.c align paremeters Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update examples/server/server.cpp remove whitespace and align brace Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update common/common.cpp Remove whitespace and align brace Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * unified ggml_numa_strategy enum and fixed text alignment in server.cpp example * Update ggml.c simplified return for platforms without NUMA support Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * removed redundant else from cli argument processing of --numa * whitespace --------- Co-authored-by: root <root@nenya.lothlorien.ca> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Jared Van Bortel <jared@nomic.ai>
2024-02-16 09:31:07 +00:00
void llama_numa_init(enum ggml_numa_strategy numa) {
if (numa != GGML_NUMA_STRATEGY_DISABLED) {
ggml_numa_init(numa);
}
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
void llama_backend_free(void) {
#ifdef GGML_USE_MPI
ggml_mpi_backend_free();
#endif
ggml : add IQ2 to test-backend-ops + refactoring (#4990) * ggml : add IQ2 to test-backend-ops + refactoring ggml-ci * cuda : update supports_op for IQ2 ggml-ci * ci : enable LLAMA_CUBLAS=1 for CUDA nodes ggml-ci * cuda : fix out-of-bounds-access in `mul_mat_vec_q` ggml-ci * tests : avoid creating RNGs for each Q tensor ggml-ci * tests : avoid creating RNGs for each tensor ggml-ci
2024-01-17 16:54:56 +00:00
ggml_quantize_free();
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
llama : support input embeddings directly (#1910) * add interface for float input * fixed inpL shape and type * add examples of input floats * add test example for embd input * fixed sampling * add free for context * fixed add end condition for generating * add examples for llava.py * add READMD for llava.py * add READMD for llava.py * add example of PandaGPT * refactor the interface and fixed the styles * add cmake build for embd-input * add cmake build for embd-input * Add MiniGPT-4 example * change the order of the args of llama_eval_internal * fix ci error
2023-06-28 15:53:37 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
int64_t llama_time_us(void) {
return ggml_time_us();
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : make model stateless and context stateful (llama_state) (#1797) * llama : make model stateless and context stateful * llama : minor cleanup * llama : update internal API declaration * Apply suggestions from code review fix style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Missing model memory release * Fix style * Add deprecated warning for public API function llama_init_from_file * Update public API use cases: move away from deprecated llama_init_from_file * Deprecate public API function llama_apply_lora_from_file --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-24 08:47:58 +00:00
struct llama_model * llama_load_model_from_file(
llama : remove LLAMA_MAX_DEVICES and LLAMA_SUPPORTS_GPU_OFFLOAD (#5240) * llama : remove LLAMA_MAX_DEVICES from llama.h ggml-ci * Update llama.cpp Co-authored-by: slaren <slarengh@gmail.com> * server : remove LLAMA_MAX_DEVICES ggml-ci * llama : remove LLAMA_SUPPORTS_GPU_OFFLOAD ggml-ci * train : remove LLAMA_SUPPORTS_GPU_OFFLOAD * readme : add deprecation notice * readme : change deprecation notice to "remove" and fix url * llama : remove gpu includes from llama.h ggml-ci --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-01-31 15:30:17 +00:00
const char * path_model,
struct llama_model_params params) {
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
ggml_time_init();
llama : make model stateless and context stateful (llama_state) (#1797) * llama : make model stateless and context stateful * llama : minor cleanup * llama : update internal API declaration * Apply suggestions from code review fix style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Missing model memory release * Fix style * Add deprecated warning for public API function llama_init_from_file * Update public API use cases: move away from deprecated llama_init_from_file * Deprecate public API function llama_apply_lora_from_file --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-24 08:47:58 +00:00
llama_model * model = new llama_model;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
unsigned cur_percentage = 0;
if (params.progress_callback == NULL) {
params.progress_callback_user_data = &cur_percentage;
params.progress_callback = [](float progress, void * ctx) {
unsigned * cur_percentage_p = (unsigned *) ctx;
unsigned percentage = (unsigned) (100 * progress);
while (percentage > *cur_percentage_p) {
*cur_percentage_p = percentage;
LLAMA_LOG_INFO(".");
if (percentage >= 100) {
LLAMA_LOG_INFO("\n");
}
}
llama : add ability to cancel model loading (#4462) * llama : Add ability to cancel model load Updated llama_progress_callback so that if it returns false, the model loading is aborted. * llama : Add test for model load cancellation * Fix bool return in llama_model_load, remove std::ignore use * Update llama.cpp Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Fail test if model file is missing * Revert "Fail test if model file is missing" This reverts commit 32ebd525bf7e5a87ee8a3dbaab3d92ce79fbf23d. * Add test-model-load-cancel to Makefile * Revert "Revert "Fail test if model file is missing"" This reverts commit 2796953257ee5383fa7c8fe8fa8fc888c048fb0b. * Simplify .gitignore for tests, clang-tidy fixes * Label all ctest tests * ci : ctest uses -L main * Attempt at writing ctest_with_model * ci : get ci/run.sh working with test-model-load-cancel * ci : restrict .github/workflows/build.yml ctest to -L main * update requirements.txt * Disable test-model-load-cancel in make * Remove venv before creation * Restructure requirements.txt Top-level now imports the specific additional requirements for each python file. Using `pip install -r requirements.txt` will fail if versions become mismatched in the per-file requirements. * Make per-python-script requirements work alone This doesn't break the main requirements.txt. * Add comment * Add convert-persimmon-to-gguf.py to new requirements.txt scheme * Add check-requirements.sh script and GitHub workflow * Remove shellcheck installation step from workflow * Add nocleanup special arg * Fix merge see: https://github.com/ggerganov/llama.cpp/pull/4462#discussion_r1434593573 * reset to upstream/master * Redo changes for cancelling model load --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2023-12-22 06:19:36 +00:00
return true;
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
};
}
llama : add ability to cancel model loading (#4462) * llama : Add ability to cancel model load Updated llama_progress_callback so that if it returns false, the model loading is aborted. * llama : Add test for model load cancellation * Fix bool return in llama_model_load, remove std::ignore use * Update llama.cpp Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> * Fail test if model file is missing * Revert "Fail test if model file is missing" This reverts commit 32ebd525bf7e5a87ee8a3dbaab3d92ce79fbf23d. * Add test-model-load-cancel to Makefile * Revert "Revert "Fail test if model file is missing"" This reverts commit 2796953257ee5383fa7c8fe8fa8fc888c048fb0b. * Simplify .gitignore for tests, clang-tidy fixes * Label all ctest tests * ci : ctest uses -L main * Attempt at writing ctest_with_model * ci : get ci/run.sh working with test-model-load-cancel * ci : restrict .github/workflows/build.yml ctest to -L main * update requirements.txt * Disable test-model-load-cancel in make * Remove venv before creation * Restructure requirements.txt Top-level now imports the specific additional requirements for each python file. Using `pip install -r requirements.txt` will fail if versions become mismatched in the per-file requirements. * Make per-python-script requirements work alone This doesn't break the main requirements.txt. * Add comment * Add convert-persimmon-to-gguf.py to new requirements.txt scheme * Add check-requirements.sh script and GitHub workflow * Remove shellcheck installation step from workflow * Add nocleanup special arg * Fix merge see: https://github.com/ggerganov/llama.cpp/pull/4462#discussion_r1434593573 * reset to upstream/master * Redo changes for cancelling model load --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com>
2023-12-22 06:19:36 +00:00
int status = llama_model_load(path_model, *model, params);
GGML_ASSERT(status <= 0);
if (status < 0) {
if (status == -1) {
LLAMA_LOG_ERROR("%s: failed to load model\n", __func__);
} else if (status == -2) {
LLAMA_LOG_INFO("%s: cancelled model load\n", __func__);
}
llama : make model stateless and context stateful (llama_state) (#1797) * llama : make model stateless and context stateful * llama : minor cleanup * llama : update internal API declaration * Apply suggestions from code review fix style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Missing model memory release * Fix style * Add deprecated warning for public API function llama_init_from_file * Update public API use cases: move away from deprecated llama_init_from_file * Deprecate public API function llama_apply_lora_from_file --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-24 08:47:58 +00:00
delete model;
return nullptr;
}
return model;
}
void llama_free_model(struct llama_model * model) {
delete model;
}
struct llama_context * llama_new_context_with_model(
ggml : change ggml_graph_compute() API to not require context (#1999) * ggml_graph_compute: deprecate using ggml_context, try resolve issue #287 * rewrite: no longer consider backward compitability; plan and make_plan * minor: rename ctx as plan; const * remove ggml_graph_compute from tests/test-grad0.c, but current change breaks backward * add static ggml_graph_compute_sugar() * minor: update comments * reusable buffers * ggml : more consistent naming + metal fixes * ggml : fix docs * tests : disable grad / opt + minor naming changes * ggml : add ggml_graph_compute_with_ctx() - backwards compatible API - deduplicates a lot of copy-paste * ci : enable test-grad0 * examples : factor out plan allocation into a helper function * llama : factor out plan stuff into a helper function * ci : fix env * llama : fix duplicate symbols + refactor example benchmark * ggml : remove obsolete assert + refactor n_tasks section * ggml : fix indentation in switch * llama : avoid unnecessary bool * ggml : remove comments from source file and match order in header --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-07 16:24:01 +00:00
struct llama_model * model,
struct llama_context_params params) {
llama : make model stateless and context stateful (llama_state) (#1797) * llama : make model stateless and context stateful * llama : minor cleanup * llama : update internal API declaration * Apply suggestions from code review fix style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Missing model memory release * Fix style * Add deprecated warning for public API function llama_init_from_file * Update public API use cases: move away from deprecated llama_init_from_file * Deprecate public API function llama_apply_lora_from_file --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-24 08:47:58 +00:00
if (!model) {
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
LLAMA_LOG_ERROR("%s: model cannot be NULL\n", __func__);
return nullptr;
}
if (params.n_batch == 0 && params.n_ubatch == 0) {
LLAMA_LOG_ERROR("%s: n_batch and n_ubatch cannot both be zero\n", __func__);
return nullptr;
}
if (params.n_ctx == 0 && model->hparams.n_ctx_train == 0) {
LLAMA_LOG_ERROR("%s: n_ctx and model->hparams.n_ctx_train cannot both be zero\n", __func__);
llama : make model stateless and context stateful (llama_state) (#1797) * llama : make model stateless and context stateful * llama : minor cleanup * llama : update internal API declaration * Apply suggestions from code review fix style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Missing model memory release * Fix style * Add deprecated warning for public API function llama_init_from_file * Update public API use cases: move away from deprecated llama_init_from_file * Deprecate public API function llama_apply_lora_from_file --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-24 08:47:58 +00:00
return nullptr;
}
llama : add functions that work directly on model (#2197) * Remove vocab reference from context * Add functions that works directly with model
2023-07-14 18:55:24 +00:00
llama_context * ctx = new llama_context(*model);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
const auto & hparams = model->hparams;
auto & cparams = ctx->cparams;
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
cparams.n_seq_max = std::max(1u, params.n_seq_max);
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
cparams.n_threads = params.n_threads;
cparams.n_threads_batch = params.n_threads_batch;
cparams.yarn_ext_factor = params.yarn_ext_factor;
cparams.yarn_attn_factor = params.yarn_attn_factor;
cparams.yarn_beta_fast = params.yarn_beta_fast;
cparams.yarn_beta_slow = params.yarn_beta_slow;
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
cparams.defrag_thold = params.defrag_thold;
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
cparams.embeddings = params.embeddings;
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
cparams.offload_kqv = params.offload_kqv;
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cparams.flash_attn = params.flash_attn;
llama : allow for user specified embedding pooling type (#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-03 10:40:27 +00:00
cparams.pooling_type = params.pooling_type;
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
cparams.n_ctx = params.n_ctx == 0 ? hparams.n_ctx_train : params.n_ctx;
cparams.rope_freq_base = params.rope_freq_base == 0.0f ? hparams.rope_freq_base_train : params.rope_freq_base;
cparams.rope_freq_scale = params.rope_freq_scale == 0.0f ? hparams.rope_freq_scale_train : params.rope_freq_scale;
metal : pad n_ctx by 32 (#6177) * metal : require ne00 >= 128 for mat-mat kernels ggml-ci * llama : pad n_ctx by 32 ggml-ci
2024-03-22 07:36:03 +00:00
// this is necessary due to kv_self.n being padded later during inference
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
cparams.n_ctx = GGML_PAD(cparams.n_ctx, 256);
metal : pad n_ctx by 32 (#6177) * metal : require ne00 >= 128 for mat-mat kernels ggml-ci * llama : pad n_ctx by 32 ggml-ci
2024-03-22 07:36:03 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// with causal attention, the batch size is limited by the context size
cparams.n_batch = hparams.causal_attn ? std::min(cparams.n_ctx, params.n_batch) : params.n_batch;
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
// the batch has to be at least GGML_KQ_MASK_PAD because we will be padding the KQ_mask
// this is required by GPU kernels in order to avoid out-of-bounds accesses (e.g. ggml_flash_attn_ext)
// ref: https://github.com/ggerganov/llama.cpp/pull/5021
if (cparams.n_batch < GGML_KQ_MASK_PAD) {
LLAMA_LOG_WARN("%s: n_batch is less than GGML_KQ_MASK_PAD - increasing to %d\n", __func__, GGML_KQ_MASK_PAD);
cparams.n_batch = GGML_KQ_MASK_PAD;
}
cparams.n_ubatch = std::min(cparams.n_batch, params.n_ubatch == 0 ? params.n_batch : params.n_ubatch);
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
cparams.n_yarn_orig_ctx = params.yarn_orig_ctx != 0 ? params.yarn_orig_ctx :
hparams.n_yarn_orig_ctx != 0 ? hparams.n_yarn_orig_ctx :
hparams.n_ctx_train;
backend : add eval callback (#4935) * backend : add eval callback ggml-ci * backend : group nodes in a single compute when user don't need them * backend : clean-up the implementation ggml-ci * simple : do not perform tensor data copy if not needed * simple : fix * simple : no need for ggml_is_contiguous + fix bool parse * llama : fix callback placement in llama_context_params * backend : avoid double-ask callback calls * simple : restore examples, imatrix will serve as a demo
2024-01-17 16:39:41 +00:00
cparams.cb_eval = params.cb_eval;
cparams.cb_eval_user_data = params.cb_eval_user_data;
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
auto rope_scaling_type = params.rope_scaling_type;
code : normalize enum names (#5697) * coda : normalize enum names ggml-ci * code : cont * code : cont
2024-02-25 10:09:09 +00:00
if (rope_scaling_type == LLAMA_ROPE_SCALING_TYPE_UNSPECIFIED) {
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
rope_scaling_type = hparams.rope_scaling_type_train;
}
code : normalize enum names (#5697) * coda : normalize enum names ggml-ci * code : cont * code : cont
2024-02-25 10:09:09 +00:00
if (rope_scaling_type == LLAMA_ROPE_SCALING_TYPE_NONE) {
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
cparams.rope_freq_scale = 1.0f; // never scale if scaling type is none
}
llama : change yarn_ext_factor placeholder to -1 (#3922)
2023-11-03 06:31:58 +00:00
if (cparams.yarn_ext_factor < 0.0f) { // negative indicates 'not set'
code : normalize enum names (#5697) * coda : normalize enum names ggml-ci * code : cont * code : cont
2024-02-25 10:09:09 +00:00
cparams.yarn_ext_factor = rope_scaling_type == LLAMA_ROPE_SCALING_TYPE_YARN ? 1.0f : 0.0f;
llama : implement YaRN RoPE scaling (#2268) Co-authored-by: cebtenzzre <cebtenzzre@gmail.com> Co-authored-by: Jeffrey Quesnelle <jquesnelle@gmail.com>
2023-11-01 22:04:33 +00:00
}
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
llama : add support for GritLM (#5959) * add gritlm example * gritlm results match * tabs to spaces * comment out debug printing * rebase to new embed * gritlm embeddings are back babeee * add to gitignore * allow to toggle embedding mode * Clean-up GritLM sample code. * Fix types. * Flush stdout and output ending newline if streaming. * mostly style fixes; correct KQ_mask comment * add causal_attn flag to llama_cparams * gritml : minor * llama : minor --------- Co-authored-by: Douglas Hanley <thesecretaryofwar@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-10 15:56:30 +00:00
cparams.causal_attn = hparams.causal_attn;
llama : allow for user specified embedding pooling type (#5849) * allow for user specified pooling type * llama : use enum types over int --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-03 10:40:27 +00:00
if (cparams.pooling_type == LLAMA_POOLING_TYPE_UNSPECIFIED) {
if (hparams.pooling_type == LLAMA_POOLING_TYPE_UNSPECIFIED) {
cparams.pooling_type = LLAMA_POOLING_TYPE_NONE;
} else {
cparams.pooling_type = hparams.pooling_type;
}
}
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
if (cparams.flash_attn && hparams.use_alibi) {
LLAMA_LOG_WARN("%s: flash_attn is not yet compatible with ALiBi - forcing off\n", __func__);
cparams.flash_attn = false;
}
if (cparams.flash_attn && model->arch == LLM_ARCH_GROK) {
LLAMA_LOG_WARN("%s: flash_attn is not compatible with Grok - forcing off\n", __func__);
cparams.flash_attn = false;
}
#ifdef GGML_USE_HIPBLAS
if (cparams.flash_attn) {
LLAMA_LOG_WARN("%s: flash_attn is not yet compatible with HIPBLAS builds - forcing off\n", __func__);
cparams.flash_attn = false;
}
#endif
Use unsigned for random seed (#2006) * Use unsigned for random seed. Keep -1 as the value to use a time based seed. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-29 13:15:15 +00:00
if (params.seed == LLAMA_DEFAULT_SEED) {
Init llama_context_params properly from CLI (#370)
2023-03-22 05:45:00 +00:00
params.seed = time(NULL);
}
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
LLAMA_LOG_INFO("%s: n_ctx = %u\n", __func__, cparams.n_ctx);
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
LLAMA_LOG_INFO("%s: n_batch = %u\n", __func__, cparams.n_batch);
LLAMA_LOG_INFO("%s: n_ubatch = %u\n", __func__, cparams.n_ubatch);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
LLAMA_LOG_INFO("%s: flash_attn = %d\n", __func__, cparams.flash_attn);
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
LLAMA_LOG_INFO("%s: freq_base = %.1f\n", __func__, cparams.rope_freq_base);
LLAMA_LOG_INFO("%s: freq_scale = %g\n", __func__, cparams.rope_freq_scale);
llama : add abort_callback to interrupt computation (#5409) * using abort_callback from ggml to stop llama computation * format fix * a brief explaining comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-02 19:52:25 +00:00
ctx->abort_callback = params.abort_callback;
ctx->abort_callback_data = params.abort_callback_data;
ctx->rng = std::mt19937(params.seed);
ctx->logits_all = params.logits_all;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
uint32_t kv_size = cparams.n_ctx;
ggml_type type_k = params.type_k;
ggml_type type_v = params.type_v;
// Mamba only needs a constant number of KV cache cells per sequence
if (model->arch == LLM_ARCH_MAMBA) {
// Mamba needs at least as many KV cells as there are sequences kept at any time
llama : more consistent names of count variables (#5994) * llama : more consistent names of count variables ggml-ci * llama : n_parallel -> n_seq_max * common : fix param name * examples : fix param name
2024-03-11 15:49:47 +00:00
kv_size = std::max((uint32_t) 1, params.n_seq_max);
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
// it's probably best to keep as much precision as possible for the states
type_k = GGML_TYPE_F32; // required by ggml_ssm_conv for Mamba's conv_states
type_v = GGML_TYPE_F32; // required by ggml_ssm_scan for Mamba's ssm_states
}
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
llama : differentiate the KV dims in the attention (#4657) * Add n_key_dim and n_value_dim Some models use values that are not derived from `n_embd`. Also remove `n_embd_head` and `n_embd_gqa` because it is not clear which "head" is referred to (key or value). Fix issue #4648. * Fix `llm_build_kqv` to use `n_value_gqa` * Rebase * Rename variables * Fix llm_build_kqv to be more generic wrt n_embd_head_k * Update default values for n_embd_head_k and n_embd_head_v Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Fix llm_load_tensors: the asserts were not backcompat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-02 11:51:28 +00:00
GGML_ASSERT(hparams.n_embd_head_k % ggml_blck_size(type_k) == 0);
GGML_ASSERT(hparams.n_embd_head_v % ggml_blck_size(type_v) == 0);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
if (!hparams.vocab_only) {
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// initialize backends
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
#ifdef GGML_USE_METAL
if (model->n_gpu_layers > 0) {
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ctx->backend_metal = ggml_backend_metal_init();
if (ctx->backend_metal == nullptr) {
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
LLAMA_LOG_ERROR("%s: failed to initialize Metal backend\n", __func__);
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
llama_free(ctx);
return nullptr;
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ctx->backends.push_back(ctx->backend_metal);
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
cuda : rename build flag to LLAMA_CUDA (#6299)
2024-03-26 00:16:01 +00:00
#elif defined(GGML_USE_CUDA)
backend : offload large batches to GPU (#6083) * backend : offload large batches to GPU * fix hip * code cleanup * fix CUDA split buffers * Update ggml-backend-impl.h Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix memset without set_device * imatrix : remove sched affix from weight names * sched : add a new split if the current one has too many inputs reduce max inputs per split more cleanup * update backends ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-03-18 10:03:04 +00:00
if (model->split_mode == LLAMA_SPLIT_MODE_NONE || model->split_mode == LLAMA_SPLIT_MODE_ROW) {
code : normalize enum names (#5697) * coda : normalize enum names ggml-ci * code : cont * code : cont
2024-02-25 10:09:09 +00:00
// with split_mode LLAMA_SPLIT_MODE_NONE or LLAMA_SPLIT_MODE_ROW, only the main GPU backend is used
backend : offload large batches to GPU (#6083) * backend : offload large batches to GPU * fix hip * code cleanup * fix CUDA split buffers * Update ggml-backend-impl.h Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix memset without set_device * imatrix : remove sched affix from weight names * sched : add a new split if the current one has too many inputs reduce max inputs per split more cleanup * update backends ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-03-18 10:03:04 +00:00
ggml_backend_t backend = ggml_backend_cuda_init(model->main_gpu);
if (backend == nullptr) {
LLAMA_LOG_ERROR("%s: failed to initialize CUDA%d backend\n", __func__, model->main_gpu);
llama_free(ctx);
return nullptr;
}
ctx->backends.push_back(backend);
} else {
// LLAMA_SPLIT_MODE_LAYER requires a backend for each GPU
for (int device = 0; device < ggml_backend_cuda_get_device_count(); ++device) {
ggml_backend_t backend = ggml_backend_cuda_init(device);
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
if (backend == nullptr) {
backend : offload large batches to GPU (#6083) * backend : offload large batches to GPU * fix hip * code cleanup * fix CUDA split buffers * Update ggml-backend-impl.h Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix memset without set_device * imatrix : remove sched affix from weight names * sched : add a new split if the current one has too many inputs reduce max inputs per split more cleanup * update backends ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-03-18 10:03:04 +00:00
LLAMA_LOG_ERROR("%s: failed to initialize CUDA%d backend\n", __func__, device);
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
llama_free(ctx);
return nullptr;
}
ctx->backends.push_back(backend);
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 17:03:59 +00:00
#elif defined(GGML_USE_VULKAN)
Vulkan k-quant mmq and ggml-backend offload functionality (#6155) * Fix Vulkan no kv offload incoherence * Add k-quant mul mat mat shaders * Rework working buffer allocation, reduces vram use noticeably Clean up cpu assist code, replaced with ggml-backend offload function * Default to all dedicated GPUs * Add fallback for integrated GPUs if no dedicated GPUs are found * Add debug info which device is allocating memory * Fix Intel dequant issue Fix validation issue * Fix Vulkan GGML_OP_GET_ROWS implementation * Clean up merge artifacts * Remove Vulkan warning
2024-03-29 16:29:21 +00:00
if (model->split_mode == LLAMA_SPLIT_MODE_ROW) {
LLAMA_LOG_ERROR("%s: Row split not supported. Failed to initialize Vulkan backend\n", __func__);
llama_free(ctx);
return nullptr;
}
if (model->split_mode == LLAMA_SPLIT_MODE_NONE) {
ggml_backend_t backend = ggml_backend_vk_init(0);
if (backend == nullptr) {
LLAMA_LOG_ERROR("%s: failed to initialize Vulkan backend\n", __func__);
llama_free(ctx);
return nullptr;
}
ctx->backends.push_back(backend);
} else {
Basic Vulkan Multi-GPU implementation (#5321) * Initial Vulkan multi-gpu implementation Move most global variables into backend context * Add names to backend device functions * Add further missing cleanup code * Reduce code duplication in tensor split layer assignment * generalize LLAMA_SPLIT_LAYER for all backends, do not expose device count and memory in llama.h * Only do device info print in the beginning and initialize one backend for cpu assist Add missing cleanup code * Rework backend memory management to make sure devices and buffers get properly allocated and freed * Rename cpu assist free function --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-07 06:54:50 +00:00
for (int device = 0; device < ggml_backend_vk_get_device_count(); ++device) {
ggml_backend_t backend = ggml_backend_vk_init(device);
if (backend == nullptr) {
LLAMA_LOG_ERROR("%s: failed to initialize Vulkan%d backend\n", __func__, device);
llama_free(ctx);
return nullptr;
}
ctx->backends.push_back(backend);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 17:03:59 +00:00
}
}
ggml : add unified SYCL backend for Intel GPUs (#2690) * first update for migration * update init_cublas * add debug functio, commit all help code * step 1 * step 2 * step3 add fp16, slower 31->28 * add GGML_LIST_DEVICE function * step 5 format device and print * step6, enhance error check, remove CUDA macro, enhance device id to fix none-zero id issue * support main device is non-zero * step7 add debug for code path, rm log * step 8, rename all macro & func from cuda by sycl * fix error of select non-zero device, format device list * ren ggml-sycl.hpp -> ggml-sycl.h * clear CMAKE to rm unused lib and options * correct queue: rm dtct:get_queue * add print tensor function to debug * fix error: wrong result in 658746bb26702e50f2c59c0e4ada8e9da6010481 * summary dpct definition in one header file to replace folder:dpct * refactor device log * mv dpct definition from folder dpct to ggml-sycl.h * update readme, refactor build script * fix build with sycl * set nthread=1 when sycl, increase performance * add run script, comment debug code * add ls-sycl-device tool * add ls-sycl-device, rm unused files * rm rear space * dos2unix * Update README_sycl.md * fix return type * remove sycl version from include path * restore rm code to fix hang issue * add syc and link for sycl readme * rm original sycl code before refactor * fix code err * add know issue for pvc hang issue * enable SYCL_F16 support * align pr4766 * check for sycl blas, better performance * cleanup 1 * remove extra endif * add build&run script, clean CMakefile, update guide by review comments * rename macro to intel hardware * editor config format * format fixes * format fixes * editor format fix * Remove unused headers * skip build sycl tool for other code path * replace tab by space * fix blas matmul function * fix mac build * restore hip dependency * fix conflict * ren as review comments * mv internal function to .cpp file * export funciton print_sycl_devices(), mv class dpct definition to source file * update CI/action for sycl code, fix CI error of repeat/dup * fix action ID format issue * rm unused strategy * enable llama_f16 in ci * fix conflict * fix build break on MacOS, due to CI of MacOS depend on external ggml, instead of internal ggml * fix ci cases for unsupported data type * revert unrelated changed in cuda cmake remove useless nommq fix typo of GGML_USE_CLBLAS_SYCL * revert hip cmake changes * fix indent * add prefix in func name * revert no mmq * rm cpu blas duplicate * fix no_new_line * fix src1->type==F16 bug. * pass batch offset for F16 src1 * fix batch error * fix wrong code * revert sycl checking in test-sampling * pass void as arguments of ggml_backend_sycl_print_sycl_devices * remove extra blank line in test-sampling * revert setting n_threads in sycl * implement std::isinf for icpx with fast math. * Update ci/run.sh Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update examples/sycl/run-llama2.sh Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update examples/sycl/run-llama2.sh Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * add copyright and MIT license declare * update the cmd example --------- Co-authored-by: jianyuzh <jianyu.zhang@intel.com> Co-authored-by: luoyu-intel <yu.luo@intel.com> Co-authored-by: Meng, Hengyu <hengyu.meng@intel.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 15:56:23 +00:00
#elif defined(GGML_USE_SYCL)
[SYCL] offload op (#6217) * remove no USM methods * leave the schedule to ggml_backend_sched entirely
2024-03-24 04:04:25 +00:00
// with split_mode LLAMA_SPLIT_MODE_NONE or LLAMA_SPLIT_MODE_ROW, only the main GPU backend is used
if (model->split_mode == LLAMA_SPLIT_MODE_NONE || model->split_mode == LLAMA_SPLIT_MODE_ROW) {
ggml_backend_t backend = ggml_backend_sycl_init(model->main_gpu);
if (backend == nullptr) {
int main_gpu_id = ggml_backend_sycl_get_device_id(model->main_gpu);
LLAMA_LOG_ERROR("%s: failed to initialize SYCL%d (index %d) backend\n", __func__, main_gpu_id, model->main_gpu);
llama_free(ctx);
return nullptr;
}
ctx->backends.push_back(backend);
} else {
// LLAMA_SPLIT_LAYER requires a backend for each GPU
for (int i = 0; i < ggml_backend_sycl_get_device_count(); ++i) {
ggml_backend_t backend = ggml_backend_sycl_init(i);
Support multiple GPUs (split mode) on SYCL backend (#5806) * suport multiple cards: split-mode - layer|row * rm warning * rebase with master, support tow new OPs, close feature for -sm=row, fix for unit test * update news * fix merge error * update according to review comments
2024-03-02 11:49:30 +00:00
if (backend == nullptr) {
[SYCL] offload op (#6217) * remove no USM methods * leave the schedule to ggml_backend_sched entirely
2024-03-24 04:04:25 +00:00
int id_list[GGML_SYCL_MAX_DEVICES];
ggml_sycl_get_gpu_list(id_list, GGML_SYCL_MAX_DEVICES);
LLAMA_LOG_ERROR("%s: failed to initialize SYCL%d (index %d) backend\n", __func__, id_list[i], i);
Support multiple GPUs (split mode) on SYCL backend (#5806) * suport multiple cards: split-mode - layer|row * rm warning * rebase with master, support tow new OPs, close feature for -sm=row, fix for unit test * update news * fix merge error * update according to review comments
2024-03-02 11:49:30 +00:00
llama_free(ctx);
return nullptr;
}
ctx->backends.push_back(backend);
ggml : add unified SYCL backend for Intel GPUs (#2690) * first update for migration * update init_cublas * add debug functio, commit all help code * step 1 * step 2 * step3 add fp16, slower 31->28 * add GGML_LIST_DEVICE function * step 5 format device and print * step6, enhance error check, remove CUDA macro, enhance device id to fix none-zero id issue * support main device is non-zero * step7 add debug for code path, rm log * step 8, rename all macro & func from cuda by sycl * fix error of select non-zero device, format device list * ren ggml-sycl.hpp -> ggml-sycl.h * clear CMAKE to rm unused lib and options * correct queue: rm dtct:get_queue * add print tensor function to debug * fix error: wrong result in 658746bb26702e50f2c59c0e4ada8e9da6010481 * summary dpct definition in one header file to replace folder:dpct * refactor device log * mv dpct definition from folder dpct to ggml-sycl.h * update readme, refactor build script * fix build with sycl * set nthread=1 when sycl, increase performance * add run script, comment debug code * add ls-sycl-device tool * add ls-sycl-device, rm unused files * rm rear space * dos2unix * Update README_sycl.md * fix return type * remove sycl version from include path * restore rm code to fix hang issue * add syc and link for sycl readme * rm original sycl code before refactor * fix code err * add know issue for pvc hang issue * enable SYCL_F16 support * align pr4766 * check for sycl blas, better performance * cleanup 1 * remove extra endif * add build&run script, clean CMakefile, update guide by review comments * rename macro to intel hardware * editor config format * format fixes * format fixes * editor format fix * Remove unused headers * skip build sycl tool for other code path * replace tab by space * fix blas matmul function * fix mac build * restore hip dependency * fix conflict * ren as review comments * mv internal function to .cpp file * export funciton print_sycl_devices(), mv class dpct definition to source file * update CI/action for sycl code, fix CI error of repeat/dup * fix action ID format issue * rm unused strategy * enable llama_f16 in ci * fix conflict * fix build break on MacOS, due to CI of MacOS depend on external ggml, instead of internal ggml * fix ci cases for unsupported data type * revert unrelated changed in cuda cmake remove useless nommq fix typo of GGML_USE_CLBLAS_SYCL * revert hip cmake changes * fix indent * add prefix in func name * revert no mmq * rm cpu blas duplicate * fix no_new_line * fix src1->type==F16 bug. * pass batch offset for F16 src1 * fix batch error * fix wrong code * revert sycl checking in test-sampling * pass void as arguments of ggml_backend_sycl_print_sycl_devices * remove extra blank line in test-sampling * revert setting n_threads in sycl * implement std::isinf for icpx with fast math. * Update ci/run.sh Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update examples/sycl/run-llama2.sh Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update examples/sycl/run-llama2.sh Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update CMakeLists.txt Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * add copyright and MIT license declare * update the cmd example --------- Co-authored-by: jianyuzh <jianyu.zhang@intel.com> Co-authored-by: luoyu-intel <yu.luo@intel.com> Co-authored-by: Meng, Hengyu <hengyu.meng@intel.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 15:56:23 +00:00
}
}
Nomic Vulkan backend (#4456) Signed-off-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: niansa <anton-sa@web.de> Co-authored-by: Adam Treat <treat.adam@gmail.com> Co-authored-by: Aaron Miller <apage43@ninjawhale.com> Co-authored-by: ToKiNoBug <tokinobug@163.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
2024-01-29 20:50:50 +00:00
#elif defined(GGML_USE_KOMPUTE)
if (model->n_gpu_layers > 0) {
auto * backend = ggml_backend_kompute_init(model->main_gpu);
if (backend == nullptr) {
LLAMA_LOG_ERROR("%s: failed to initialize Kompute backend\n", __func__);
llama_free(ctx);
return nullptr;
}
ctx->backends.push_back(backend);
}
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
#endif
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ctx->backend_cpu = ggml_backend_cpu_init();
if (ctx->backend_cpu == nullptr) {
LLAMA_LOG_ERROR("%s: failed to initialize CPU backend\n", __func__);
llama_free(ctx);
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
return nullptr;
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ctx->backends.push_back(ctx->backend_cpu);
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
if (!llama_kv_cache_init(ctx->kv_self, ctx, type_k, type_v, kv_size, cparams.offload_kqv)) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
LLAMA_LOG_ERROR("%s: llama_kv_cache_init() failed for self-attention cache\n", __func__);
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
llama_free(ctx);
return nullptr;
}
{
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
size_t memory_size_k = 0;
size_t memory_size_v = 0;
for (auto & k : ctx->kv_self.k_l) {
memory_size_k += ggml_nbytes(k);
}
for (auto & v : ctx->kv_self.v_l) {
memory_size_v += ggml_nbytes(v);
}
LLAMA_LOG_INFO("%s: KV self size = %7.2f MiB, K (%s): %7.2f MiB, V (%s): %7.2f MiB\n", __func__,
(float)(memory_size_k + memory_size_v) / (1024.0f * 1024.0f),
ggml_type_name(type_k), (float)memory_size_k / (1024.0f * 1024.0f),
ggml_type_name(type_v), (float)memory_size_v / (1024.0f * 1024.0f));
Reduce memory usage and allocate enough memory for largest context (#473) * Reduce memory usage and allocate enough memory for large contexts * Simpler scratch buffer usage * Reenable BLAS for quantized mul_mat * Fix number of layers in 30B and 65B * Fix KV cache size for F32
2023-03-24 21:17:37 +00:00
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// graph outputs buffer
{
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
// resized during inference when a batch uses more outputs
if (llama_output_reserve(*ctx, params.n_seq_max) < params.n_seq_max) {
LLAMA_LOG_ERROR("%s: failed to reserve initial output buffer\n", __func__);
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
llama_free(ctx);
return nullptr;
}
LLAMA_LOG_INFO("%s: %10s output buffer size = %8.2f MiB\n", __func__,
ggml_backend_buffer_name(ctx->buf_output),
ggml_backend_buffer_get_size(ctx->buf_output) / 1024.0 / 1024.0);
llama : pre-allocate input tensors in a separate buffer (#5100)
2024-01-24 11:48:14 +00:00
}
// scheduler and compute buffers
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
{
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// buffer types used for the compute buffer of each backend
std::vector<ggml_backend_buffer_type_t> backend_buft;
for (auto * backend : ctx->backends) {
if (ggml_backend_is_cpu(backend)) {
// use host buffers for the CPU backend compute buffer
backend_buft.push_back(llama_default_buffer_type_cpu(true));
} else {
backend_buft.push_back(ggml_backend_get_default_buffer_type(backend));
}
}
// buffer used to store the computation graph and the tensor meta data
llama : fix defrag bugs + add parameter (#5735) * llama : fix defrag bugs + enable by default ggml-ci * llama : add defrag_thold parameter ggml-ci * llama : cont * llama : disable log message ggml-ci * llama : fix graph size check during defrag
2024-02-27 12:35:51 +00:00
ctx->buf_compute_meta.resize(ggml_tensor_overhead()*LLAMA_MAX_NODES + ggml_graph_overhead_custom(LLAMA_MAX_NODES, false));
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// enabling pipeline parallelism in the scheduler increases memory usage, so it is only done when necessary
bool pipeline_parallel = llama_get_device_count() > 1 && model->n_gpu_layers > (int)model->hparams.n_layer && model->split_mode == LLAMA_SPLIT_MODE_LAYER;
cuda : rename build flag to LLAMA_CUDA (#6299)
2024-03-26 00:16:01 +00:00
#ifndef GGML_USE_CUDA
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
// pipeline parallelism requires support for async compute and events
// currently this is only implemented in the CUDA backend
pipeline_parallel = false;
#endif
ctx->sched = ggml_backend_sched_new(ctx->backends.data(), backend_buft.data(), ctx->backends.size(), LLAMA_MAX_NODES, pipeline_parallel);
if (pipeline_parallel) {
LLAMA_LOG_INFO("%s: pipeline parallelism enabled (n_copies=%d)\n", __func__, ggml_backend_sched_get_n_copies(ctx->sched));
}
ggml : add graph tensor allocator (#2411) * ggml : add graph tensor allocator * ggml : don't calculate data pointer of unallocated tensors when creating a view with an offset * ggml : refactor ggml_view_Nd into ggml_view_tensor_offset
2023-07-30 13:58:01 +00:00
// build worst-case graph
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
int n_tokens = (int)std::min(cparams.n_ctx, cparams.n_ubatch);
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
int n_past = cparams.n_ctx - n_tokens;
llama : remove token functions with `context` args in favor of `model` (#3720) * added `llama_model_token_*` variants to all the `llama_token_*` functions. * added `LLAMA_API` * formatting Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * removed old `llama_token` functions * changed 3 more functions to take in model - `llama_token_get_text` - `llama_token_get_score` - `llama_token_get_type` * added back docs * fixed main.cpp * changed token functions to use new model variants * changed token functions to use new model variants --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-23 19:40:03 +00:00
llama_token token = llama_token_bos(&ctx->model); // not actually used by llama_build_graph, but required to choose between token and embedding inputs graph
sync : ggml (#5452) * ggml-alloc : v3 (ggml/727) * ggml-alloc v3 ggml-ci * fix ci ggml-ci * whisper : check for backend buffer allocation failures * whisper : avoid leaks when initialization fails * cleanup ggml-ci * style fixes ggml-ci * sync : ggml * update llama.cpp, clip.cpp, export-lora.cpp * update finetune.cpp, train-text-from-scratch.cpp ggml-ci * ggml-backend : reduce alignment to 32 to match gguf and fix mmap --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-12 07:16:06 +00:00
ggml_cgraph * gf = llama_build_graph(*ctx, llama_batch_get_one(&token, n_tokens, n_past, 0), true);
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// initialize scheduler with the worst-case graph
sync : ggml (#5452) * ggml-alloc : v3 (ggml/727) * ggml-alloc v3 ggml-ci * fix ci ggml-ci * whisper : check for backend buffer allocation failures * whisper : avoid leaks when initialization fails * cleanup ggml-ci * style fixes ggml-ci * sync : ggml * update llama.cpp, clip.cpp, export-lora.cpp * update finetune.cpp, train-text-from-scratch.cpp ggml-ci * ggml-backend : reduce alignment to 32 to match gguf and fix mmap --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-12 07:16:06 +00:00
if (!ggml_backend_sched_reserve(ctx->sched, gf)) {
LLAMA_LOG_ERROR("%s: failed to allocate compute buffers\n", __func__);
llama_free(ctx);
return nullptr;
}
CUDA GPU acceleration for LoRAs + f16 models (#1970)
2023-06-28 16:35:54 +00:00
sync : ggml (#5452) * ggml-alloc : v3 (ggml/727) * ggml-alloc v3 ggml-ci * fix ci ggml-ci * whisper : check for backend buffer allocation failures * whisper : avoid leaks when initialization fails * cleanup ggml-ci * style fixes ggml-ci * sync : ggml * update llama.cpp, clip.cpp, export-lora.cpp * update finetune.cpp, train-text-from-scratch.cpp ggml-ci * ggml-backend : reduce alignment to 32 to match gguf and fix mmap --------- Co-authored-by: slaren <slarengh@gmail.com>
2024-02-12 07:16:06 +00:00
for (size_t i = 0; i < ctx->backends.size(); i++) {
ggml_backend_t backend = ctx->backends[i];
ggml_backend_buffer_type_t buft = backend_buft[i];
size_t size = ggml_backend_sched_get_buffer_size(ctx->sched, backend);
backend : offload large batches to GPU (#6083) * backend : offload large batches to GPU * fix hip * code cleanup * fix CUDA split buffers * Update ggml-backend-impl.h Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix memset without set_device * imatrix : remove sched affix from weight names * sched : add a new split if the current one has too many inputs reduce max inputs per split more cleanup * update backends ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-03-18 10:03:04 +00:00
if (size > 1) {
LLAMA_LOG_INFO("%s: %10s compute buffer size = %8.2f MiB\n", __func__,
ggml_backend_buft_name(buft),
size / 1024.0 / 1024.0);
}
build : on Mac OS enable Metal by default (#2901) * build : on Mac OS enable Metal by default * make : try to fix build on Linux * make : move targets back to the top * make : fix target clean * llama : enable GPU inference by default with Metal * llama : fix vocab_only logic when GPU is enabled * common : better `n_gpu_layers` assignment * readme : update Metal instructions * make : fix merge conflict remnants * gitignore : metal
2023-09-04 19:26:24 +00:00
}
llama : pre-allocate input tensors in a separate buffer (#5100)
2024-01-24 11:48:14 +00:00
// note: the number of splits during measure is higher than during inference due to the kv shift
int n_splits = ggml_backend_sched_get_n_splits(ctx->sched);
backend : offload large batches to GPU (#6083) * backend : offload large batches to GPU * fix hip * code cleanup * fix CUDA split buffers * Update ggml-backend-impl.h Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix memset without set_device * imatrix : remove sched affix from weight names * sched : add a new split if the current one has too many inputs reduce max inputs per split more cleanup * update backends ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-03-18 10:03:04 +00:00
LLAMA_LOG_INFO("%s: graph nodes = %d\n", __func__, gf->n_nodes);
LLAMA_LOG_INFO("%s: graph splits = %d\n", __func__, n_splits);
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
build : on Mac OS enable Metal by default (#2901) * build : on Mac OS enable Metal by default * make : try to fix build on Linux * make : move targets back to the top * make : fix target clean * llama : enable GPU inference by default with Metal * llama : fix vocab_only logic when GPU is enabled * common : better `n_gpu_layers` assignment * readme : update Metal instructions * make : fix merge conflict remnants * gitignore : metal
2023-09-04 19:26:24 +00:00
}
CUDA GPU acceleration for LoRAs + f16 models (#1970)
2023-06-28 16:35:54 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
#ifdef GGML_USE_MPI
ctx->ctx_mpi = ggml_mpi_init();
Add LoRA support (#820)
2023-04-17 15:28:55 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
if (ggml_mpi_rank(ctx->ctx_mpi) > 0) {
// Enter a blocking eval loop with dummy input, letting rank=0 drive the process
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
// TODO: needs fix after #3228
GGML_ASSERT(false && "not implemented");
//const std::vector<llama_token> tmp(ctx->model.hparams.n_ctx, llama_token_bos(ctx));
//while (!llama_eval(ctx, tmp.data(), tmp.size(), 0, 0)) {};
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama_backend_free();
exit(1);
}
#endif
Add LoRA support (#820)
2023-04-17 15:28:55 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
return ctx;
}
Add LoRA support (#820)
2023-04-17 15:28:55 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
void llama_free(struct llama_context * ctx) {
delete ctx;
}
CUDA GPU acceleration for LoRAs + f16 models (#1970)
2023-06-28 16:35:54 +00:00
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
const llama_model * llama_get_model(const struct llama_context * ctx) {
return &ctx->model;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
Add LoRA support (#820)
2023-04-17 15:28:55 +00:00
llama : allow getting n_batch from llama_context in c api (#4540) * allowed getting n_batch from llama_context in c api * changed to use `uint32_t` instead of `int` * changed to use `uint32_t` instead of `int` in `llama_n_ctx` * Update llama.h --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 19:57:48 +00:00
uint32_t llama_n_ctx(const struct llama_context * ctx) {
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
return ctx->cparams.n_ctx;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
ggml : change ggml_graph_compute() API to not require context (#1999) * ggml_graph_compute: deprecate using ggml_context, try resolve issue #287 * rewrite: no longer consider backward compitability; plan and make_plan * minor: rename ctx as plan; const * remove ggml_graph_compute from tests/test-grad0.c, but current change breaks backward * add static ggml_graph_compute_sugar() * minor: update comments * reusable buffers * ggml : more consistent naming + metal fixes * ggml : fix docs * tests : disable grad / opt + minor naming changes * ggml : add ggml_graph_compute_with_ctx() - backwards compatible API - deduplicates a lot of copy-paste * ci : enable test-grad0 * examples : factor out plan allocation into a helper function * llama : factor out plan stuff into a helper function * ci : fix env * llama : fix duplicate symbols + refactor example benchmark * ggml : remove obsolete assert + refactor n_tasks section * ggml : fix indentation in switch * llama : avoid unnecessary bool * ggml : remove comments from source file and match order in header --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-07 16:24:01 +00:00
llama : allow getting n_batch from llama_context in c api (#4540) * allowed getting n_batch from llama_context in c api * changed to use `uint32_t` instead of `int` * changed to use `uint32_t` instead of `int` in `llama_n_ctx` * Update llama.h --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 19:57:48 +00:00
uint32_t llama_n_batch(const struct llama_context * ctx) {
return ctx->cparams.n_batch;
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
uint32_t llama_n_ubatch(const struct llama_context * ctx) {
return ctx->cparams.n_ubatch;
}
llama : more consistent names of count variables (#5994) * llama : more consistent names of count variables ggml-ci * llama : n_parallel -> n_seq_max * common : fix param name * examples : fix param name
2024-03-11 15:49:47 +00:00
uint32_t llama_n_seq_max(const struct llama_context * ctx) {
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
return ctx->kv_self.size;
}
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
enum llama_vocab_type llama_vocab_type(const struct llama_model * model) {
return model->vocab.type;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
Add LoRA support (#820)
2023-04-17 15:28:55 +00:00
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
enum llama_rope_type llama_rope_type(const struct llama_model * model) {
switch (model->arch) {
// these models do not use RoPE
case LLM_ARCH_GPT2:
case LLM_ARCH_GPTJ:
case LLM_ARCH_GPTNEOX:
case LLM_ARCH_MPT:
case LLM_ARCH_REFACT:
case LLM_ARCH_BLOOM:
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
case LLM_ARCH_MAMBA:
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
return LLAMA_ROPE_TYPE_NONE;
// use what we call a normal RoPE, operating on pairs of consecutive head values
case LLM_ARCH_LLAMA:
case LLM_ARCH_BAICHUAN:
case LLM_ARCH_STARCODER:
case LLM_ARCH_PLAMO:
case LLM_ARCH_CODESHELL:
case LLM_ARCH_ORION:
case LLM_ARCH_INTERNLM2:
case LLM_ARCH_MINICPM:
[Model] Add support for xverse (#6301) * Support xverse model convert to gguf format. * 1. Convert xverse models to gguf; 2. Add LLM_ARCH_XVERSE inference in llama.cpp; 3. Add xverse item in Supported models in README.md; * * gguf-py: remove redundant logs * llama: remove the init_mapping_prefetch custom parameter * llama.cpp: Include the changes from #6122 to exclude the unused outputs of the last layers. * - Fix format issues - Remove duplicate set kqv_out to llm_build_kv * Update llama.cpp --------- Co-authored-by: willhe <willhe@xverse.cn> Co-authored-by: willhe <hexin@xverse.cn>
2024-03-29 13:37:03 +00:00
case LLM_ARCH_XVERSE:
llama : add Command-R support (#6033) Information about the Command-R 35B model (128k context) can be found at: https://huggingface.co/CohereForAI/c4ai-command-r-v01 Based on the llama2 model with a few changes: 1) New hyper parameter to scale output logits (logit_scale) 2) Uses LayerNorm instead of RMSNorm 3) Transfomer layers have a single shared LayerNorm that feeds into both the self-attention and FFN layers in parallel. There is no post-attention LayerNorm. 4) No support for Rotary Position Embeddings (RoPE) scaling 5) No biases used Find GGUF files here: https://huggingface.co/andrewcanis/c4ai-command-r-v01-GGUF To convert model to GGUF format yourself: 1) Download Command-R Hugging Face safetensors: git lfs install git clone https://huggingface.co/CohereForAI/c4ai-command-r-v01 2) Run: python3 convert-hf-to-gguf.py --outtype f16 ./c4ai-command-r-v01
2024-03-15 20:41:22 +00:00
case LLM_ARCH_COMMAND_R:
Implement the OLMo architecture (#6741) * implement olmo architecture * remove unused variable * remove unused moe branch * remove check for weight * remove superfluous moe, bias and rope tensors * clarified comment * fix clamp_kqv setting * remove obsolete parameter name filter
2024-04-19 09:35:54 +00:00
case LLM_ARCH_OLMO:
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
return LLAMA_ROPE_TYPE_NORM;
// the pairs of head values are offset by n_rot/2
case LLM_ARCH_FALCON:
llama : add grok-1 support (#6204) * Add support for Grok model architecture * Revert convert-hf-to-gguf to default options * Fixed f_norm_rms_eps bug * Fix whitespaces * llama : fix grok rope type * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 16:41:53 +00:00
case LLM_ARCH_GROK:
model: support arch `DbrxForCausalLM` (#6515) * model: dbrx convert to gguf #6344 * llama: support dbrx #6344 * doc: dbrx: add the model as supported * scripts: get-wikitext-2 add unzip * llama: increase maximum experts allowed * llama: factorize moe graph implementation between grok, mixtral and dbrx --------- Co-authored-by: Megha Agarwal <16129366+megha95@users.noreply.github.com>
2024-04-13 09:33:52 +00:00
case LLM_ARCH_DBRX:
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
case LLM_ARCH_PERSIMMON:
case LLM_ARCH_BERT:
case LLM_ARCH_NOMIC_BERT:
case LLM_ARCH_STABLELM:
case LLM_ARCH_QWEN:
case LLM_ARCH_QWEN2:
llama : add qwen2moe (#6074) * support qwen2moe * fix-review * metal : support unary ops for nelements % 4 != 0 * metal : require contiguousness for float4 unary kernels * metal : require contiguousness for float4 unary kernels (cont) * fix-review * names : for brevity "SHARED_EXP" -> "SHEXP" * llama : reuse build_moe_ffn() * llama : add model type name --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-16 15:40:48 +00:00
case LLM_ARCH_QWEN2MOE:
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
case LLM_ARCH_PHI2:
llama : add phi3 support (#6852) * add explicit phi3 support * add explicit phi3 support * remove unused code * convert : add BOS token * llama : match EOT token <|end|> * llama : minor / style * llama : tabs -> spaces * convert : fix lint checks --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-24 07:00:37 +00:00
case LLM_ARCH_PHI3:
llama : fix Gemma rope type (#5691)
2024-02-26 06:30:17 +00:00
case LLM_ARCH_GEMMA:
llama : add StarCoder2 support (#5795) * Add support for starcoder2 * handle rope type * skip rope freq and rotary embeddings from being serialized * resolve comments * Update llama.cpp * remove redundant changes * handle `rope-theta` * llama : change starcoder2 rope type * address comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-01 19:30:46 +00:00
case LLM_ARCH_STARCODER2:
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
return LLAMA_ROPE_TYPE_NEOX;
// all model arches should be listed explicitly here
case LLM_ARCH_UNKNOWN:
GGML_ASSERT(false && "unknown architecture");
break;
}
return LLAMA_ROPE_TYPE_NONE;
}
llama : add llama_get_pooling_type function (#6862) * add llama_get_pooling_type function * fix argument name, move with ctx funcs
2024-04-24 13:10:07 +00:00
enum llama_pooling_type llama_pooling_type(const struct llama_context * ctx) {
return ctx->cparams.pooling_type;
}
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
int32_t llama_n_vocab(const struct llama_model * model) {
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
return model->hparams.n_vocab;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
Add LoRA support (#820)
2023-04-17 15:28:55 +00:00
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
int32_t llama_n_ctx_train(const struct llama_model * model) {
examples : make n_ctx warning work again (#3066) This was broken by commit e36ecdcc ("build : on Mac OS enable Metal by default (#2901)").
2023-09-08 15:43:35 +00:00
return model->hparams.n_ctx_train;
}
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
int32_t llama_n_embd(const struct llama_model * model) {
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
return model->hparams.n_embd;
}
Add LoRA support (#820)
2023-04-17 15:28:55 +00:00
llama : add support for control vectors (#5970) * control vector api and implementation * control-vectors : minor code style updates * disable control vector when data == nullptr use -1 for disabled range (also on init) in case we ever support controlling layer 0 (embeddings) --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-15 20:43:02 +00:00
int32_t llama_n_layer(const struct llama_model * model) {
return model->hparams.n_layer;
}
llama : expose model's rope_freq_scale in the API (#3418) so it can be scaled further before creating a context.
2023-10-03 17:09:28 +00:00
float llama_rope_freq_scale_train(const struct llama_model * model) {
return model->hparams.rope_freq_scale_train;
}
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
int32_t llama_model_meta_val_str(const struct llama_model * model, const char * key, char * buf, size_t buf_size) {
llama : add functions to get the model's metadata (#4013) * llama : add functions to get the model's metadata * format -> std::to_string * better documentation
2023-11-17 15:17:37 +00:00
const auto & it = model->gguf_kv.find(key);
if (it == model->gguf_kv.end()) {
if (buf_size > 0) {
buf[0] = '\0';
}
return -1;
}
return snprintf(buf, buf_size, "%s", it->second.c_str());
}
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
int32_t llama_model_meta_count(const struct llama_model * model) {
llama : add functions to get the model's metadata (#4013) * llama : add functions to get the model's metadata * format -> std::to_string * better documentation
2023-11-17 15:17:37 +00:00
return (int)model->gguf_kv.size();
}
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
int32_t llama_model_meta_key_by_index(const struct llama_model * model, int i, char * buf, size_t buf_size) {
llama : add functions to get the model's metadata (#4013) * llama : add functions to get the model's metadata * format -> std::to_string * better documentation
2023-11-17 15:17:37 +00:00
if (i < 0 || i >= (int)model->gguf_kv.size()) {
if (buf_size > 0) {
buf[0] = '\0';
}
return -1;
}
auto it = model->gguf_kv.begin();
std::advance(it, i);
return snprintf(buf, buf_size, "%s", it->first.c_str());
}
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
int32_t llama_model_meta_val_str_by_index(const struct llama_model * model, int32_t i, char * buf, size_t buf_size) {
llama : add functions to get the model's metadata (#4013) * llama : add functions to get the model's metadata * format -> std::to_string * better documentation
2023-11-17 15:17:37 +00:00
if (i < 0 || i >= (int)model->gguf_kv.size()) {
if (buf_size > 0) {
buf[0] = '\0';
}
return -1;
}
auto it = model->gguf_kv.begin();
std::advance(it, i);
return snprintf(buf, buf_size, "%s", it->second.c_str());
}
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
int32_t llama_model_desc(const struct llama_model * model, char * buf, size_t buf_size) {
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
return snprintf(buf, buf_size, "%s %s %s",
YaRN : store rope scaling type as int32_t in memory (#5285) * YaRN : store rope scaling type as int32_t in memory * llama : store mapped names as const char *
2024-02-03 11:22:06 +00:00
llama_model_arch_name(model->arch),
llm : add Falcon support (#2717) * llama : refactor GGUF constants into static maps * llama : check if model architecture is known * llama : refactor llama_model_load_internal() * gguf : add KV constant maps * llm : read arch-specific KVs * convert : add dummy scores + types * falcon : load tensor data (CPU only) * llama : fix loading progress bar * llama : add arch member to llama_model * falcon : CPU inference working * falcon : support non-40B models * falcon : minor * llama : minor updates ggml-ci * convert-falcon-hf-to-gguf.py : fix special token mapping * llama.cpp : llama default UNK token = id 0 * llama.cpp : fix bpe tokenizer * llama.cpp : fix the fix of bpe tokenizer * ggml : pass eps to ggml_norm * metal : implement RoPE (mode = 2) + avoid ggml_repeat * ggml : ggml_repeat always creates new tensor * falcon : copy-paste self-attention from LLaMA * metal : print extra compute pipeline info * falcon : minor changes (still chasing the Metal problem) * llama.cpp : fix linefeed token * metal : fix GELU kernel numerical stability by using precise::tanh * metal : temporary workaround for the concurrency optimization bug * falcon : add CUDA offloading (#2739) * llama : better model naming and size reporting * llama : prep new tokenizer support * llama : advanced BPE tokenizer based on ggllm.cpp imlpementation * llama : remove oboslete comment ggml-ci * common : remove obsolete BPE API + disable test-tokenizer-1 * llama : revert BPE special-case in llama_byte_to_token() * cuda : add TODOs for RoPE NeoX implementation * llama : default special tokens based on vocab type * perplexity : add log for start of tokenization --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com>
2023-08-23 20:08:04 +00:00
llama_model_type_name(model->type),
llama_model_ftype_name(model->ftype).c_str());
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
Add LoRA support (#820)
2023-04-17 15:28:55 +00:00
llama-bench : add model sizes (#2771) * llama-bench : add model sizes * more compact markdown output * back to GiB * adjust column sizes
2023-08-25 13:16:19 +00:00
uint64_t llama_model_size(const struct llama_model * model) {
uint64_t size = 0;
for (const auto & it : model->tensors_by_name) {
size += ggml_nbytes(it.second);
}
return size;
}
uint64_t llama_model_n_params(const struct llama_model * model) {
uint64_t nparams = 0;
for (const auto & it : model->tensors_by_name) {
nparams += ggml_nelements(it.second);
}
return nparams;
}
train : finetune LORA (#2632) * fix track_max_mem in forward_batch_wo_cache_flash_attn_train * remove unnecessary Adam(W) optimizer tensors. reduces optimizer memory overhead from 7*modelsize to 2*modelsize. additionally allows to optimize models with more than 2^31 parameters by replacing int with int64_t. bumps training checkpoint file version, but old checkpoints can still be read. new version with less tensors is saved. * add gradient clipping to AdamW * Fix reset of unused g->nodes and g->grads to NULL * implement gradient checkpointing for training reduces memory overhead from O(n_layer) to O(sqrt(n_layer)) as explained in readme of https://github.com/cybertronai/gradient-checkpointing * remove unused compute buffer 3 * add and use function ggml_build_backward_expand to avoid stack overflows with large maximum number of nodes GGML_API void ggml_build_backward_expand(struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, bool keep); * change AdamW decay parameter to work like the torch AdamW decay parameter It is now relative to Adam learning rate `alpha*sched`. Before that it was relative to `sched` only. `alpha` being the maximum learning rate and `sched` being a scaling parameter in [0..1] * change default AdamW weight decay parameter used in training to 0.1 as used in nanoGPT * change default AdamW weight decay parameter defined in ggml to 0.0, making Adam default instead of AdamW btw: the default weight decay parameter for torch.optim.AdamW is 0.01 * bug fixes for cross entropy loss ggml_cross_entropy_loss: sums where not correctly added in workload of each thread ggml_cross_entropy_loss_back: simplify backward process, reducing numerical issues guard usage of exp f16 lookup in cross entropy by #define GGML_CROSS_ENTROPY_EXP_FP16 cross entropy loss is only used once during training, but it is quite sensitive to numerical errors introduced by exp-f16-lookup. so exp-f16-lookup for cross entropy loss is disabled by default, trading better gradients for very slightly worse runtime performance. * fix test-grad0 for cross_entropy_loss the second argument to cross_entropy_loss must sum up to 1 for each row * fix test-grad0 for soft_max dont use only sum as aggregation, because sum of softmax is always 1 -> finite differences should not work instead use sum(log(soft_max()*(1-eps)+eps)); use eps to avoid log(0) * improve finite differences of test-grad0 by using double instead of float * change cross_entropy_loss to output average over all rows this helps keeping the loss and gradients in a sane range * improve gradient checkpointing sqrt(n_layers) is only the best checkpoint step when mem size of checkpoints and mem size of layers are equal. since layers require more memory than the single-tensor-checkpoint we use, the optimal values are compute different: ``` given: n, u, v objective: minimize(a*u+b*v) where a*b=n, a>0, b>0 b=n/a minimize(a*u+v*n/a) diff(a*u+v*n/a, a) = u - (v*n/a)/a diff(a*u+v*n/a, a) == 0 u - (v*n/a)/a == 0 u == v*n/(a*a) u*a*a = v*n a*a = v*n/u a = sqrt(n*v/u) ``` this change results in more checkpoints, requiring less layers to store between checkpoints, overall improving memory usage. * disable gradient checkpointing debug output * llama : fix rope usage in train-text-from-scratch after ChatGLM change * add more training parameters: --enable-restart N Only for Adam optimizer. Enable restarts of cos-decay --disable-restart N Only for Adam optimizer. Disable restarts of cos-decay --opt-past N Number of optimization iterations to track for delta convergence test. Disabled when zero. --opt-delta N Maximum delta for delta convergence test. Disabled when <= zero. --opt-max-no-improvement N Maximum number of optimization iterations with no improvement. Disabled when <= zero. --adam-epsf N AdamW epsilon for convergence test. Disabled when <= zero. --adam-min-alpha N Adam minimum learning rate alpha, usually 0.1 * alpha * replace memcpy with reshape operation so that the graph is not cut at the input this makes it possible to store other values into the input tensor and then simply recompute the graph without rebuilding it * remove unused function argument from get_example_targets_batch * measure and print total training time * add optimization callback to ggml_opt_resume_g this callback is called before each iteration with custom data and pointer to learning schedule parameter (only used in Adam(W)). can be used for dynamic learning schedule and setting input data for batches before each iteration * use optimization callback in training allows dynamic learning schedule and different batch data for each iteration without relying on low n_iter and high n_examples parameters reduces runtime by avoiding restart of optimization function and improves training convergence by providing a different batch for each iteration * add minimum number of tensor dimensions to apply weight decay (default 2) this allows to not apply weight decay to bias parameters * rename training parameter cos-decay-alpha to cos-decay-min and clarify that adam-min-alpha also applies to warmup * fix increase of model.train_samples and model.train_tokens now that each optimizer iteration gets its own batch we need to multiply by number of opt iterations * change sampling parameters for prediction after training to defaults of common.h and clarify what is context for prediction and what are generated tokens * tighten abs error bounds for cross_entropy_loss in test-grad0 * add conditional compilation of using F16 exp in flash attention uncomment `// #define GGML_FLASH_ATTN_EXP_FP16` to enable usage of f16 exp in flash attention * tighten abs error bounds for flash_attn in test-grad0 * tighten abs error bounds for sqrt in test-grad0 * remove out-commented vectorized code of opt_adam the vectorized code might be bit faster for low number of parameters, but it had a big memory usage overhead * ggml : update ggml_rms_norm_back with configurable eps * llama training : fix ggml_rms_norm_back calls to pass configurable eps * remove trailing whitespace * add train function using automatic gradient checkpointing backward pass and allocator * in train function replace add_inplace by regular add because using add_inplace seems to result in different gradients * don't use allocate hash_map on context because the context has no_alloc=True when using memory allocator resulting in NULL data pointers * correctly clone reshape and permute operations by also cloning tensor->nb values * fix variable name and add missing type cast * terminate recursive tensor cloning when reaching tensor without src tensors * correctly clone view tensors by setting data pointers without this the checkpointing would only work when being used together with memory allocator * fix variable names * swap arguments to commutative ops to be the same as in `forward_batch_wo_cache_flash_attn` * add input tensors as checkpoints so that recursive tensor cloning of gradient checkpointing terminates on input tensors * fix variable name and add missing boolean negation * make sure some tensors are not reallocated by inserting new temporary nodes depending on them: output and parameter gradient tensors need to be available at the end of the graph execution parameter gradient tensors also need to be available before the graph execution because they are set to zero before each optimizer iteration checkpoint tensors are allocated all together to reduce memory allocator fragmentation afterwards, in addition to the temporary nodes, we also need to reset the temporary leafs * fix ASSERT to work with zero layers * add training options whether to use allocator and/or unified training function * integrate unified training function which may use memory allocator the unified training function also supports arguments whether to use flash attention and/or gradient checkpointing * format name of cloned tensors with " (clone)" suffix * set names for tensors in unified train function for easier debugging * allocate graph on context using ggml_new_graph * remove handwritten training functions * remove unused training parameters "use_scratch" and "use_unified" * remove trailing whitespace * remove unused train params: mem_compute1_gb & mem_compute2_gb mem_compute_gb is used for compute when automatic memory allocator is not enabled, otherwise it can be very small to only hold the tensor definitions mem_compute0_gb is used for automatic memory allocator (as long as measurement of max required size is not implemented) * remove unused forward_batch function * add debug asserts in ggml_allocr_alloc to some common pitfalls when using this function directly * only use ggml_allocr_alloc when tensor has NULL data and is no view * fix test when to create temporary backward graph temporary backward graph is only necessary when using checkpointing * fix memory "leak" in optimizers each iteration a new cplan with new memory for work data was allocated. now cplan creation only happens at the start of optimization, with each iteration reusing the cplan and its work data. * reverse order of for loop in ggml_build_backward_expand to save memory when using gradient checkpointing and allocator with this loop order gradient checkpointing with allocator on 16 layer model saves 13% memory; 2 layer memory it saves 2% memory. the computation results are the same * add API functions to access llama model tensors * add stub example for finetuning, based on train-text-from-scratch * move and remove code * add API functions to access remaining model parameters: mult, head and rot * first draft for LORA finetune training * remove const model and layer arguments in API functions for accessing model tensors * bug fixes to make finetune compile automatic allocator does not work yet * add debug prints for training memory improvements * fix names of lora tensors * avoid stack overflow resulting from big ggml_cgraph replace stack allocation and ggml_build_forward by ggml_new_graph in combination with ggml_build_forward_expand * replace llama API functions to get model tensors by one function to get model tensor by name LLAMA_API struct ggml_tensor * llama_get_model_tensor(struct llama_model * model, const char * name); * remove unused call to not existing llama_get_layer_from_model * implement ggml_compute_forward_out_prod_q_f32 * remove trailing whitespace * add lora finetune support on quantized base model tensors * add ggml_add_cast API function this function works like ggml_add, but accepts a data type for the resulting tensor. only supported for quantized src0 input. * use ggml_add_cast in finetuning lora-applied weights will now have data type F32, which improves gradients when finetuning quantized base models * bug fix: actually use result type passed to ggml_add_cast * make sure base model tensors data cannot be used in viewable operations memory allocator would try to make lora application inplace on base model tensors. since those are memory mapped this will result in memory access violations * fix bug in ggml_out_prod which resulted in wrong n_dims of result tensors * avoid keeping in memory ALL of the gradients The problem here stems from ggml_graph_reset. This function is called in the optimization function, before each graph computation, to reset the gradients to zero. This required a unique memory slot for each gradient: allocating memory from a previosly freed memory location might lead to non-zero input gradients. During ggml_compute_backward the gradients are build stepwise by adding or substracting new values, starting from a OP_NONE tensor which needs to contain zero-values. This requires the graph reset. To avoid this I now remember in ggml_build_backward_expand the original OP_NONE gradient tensors in a hash table, which is passed to ggml_compute_backward. There instead of using add (or sub or similar) I test whether the existing gradient to be changed is a zero-valued-tensor by looking up its existence in the hash table. When it is such a zero-tensor it will not be modified, but replaced by the value to be added, otherwise the regular add (not inplace, allocator will take care of this) will be used. This way none of those zero-tensor values will be necessary in the final backward graph and more importantly they won't need a unique memory slot, just to make them zero. * remove trailing whitespace * remove debug prints and function to compute tensor data hash * improve optimization iteration prints * adjust maximal values to support finetuning 3B models * change default finetune params lora_r and lora_alpha to match the n_rank parameters of 4 * bug fix: make sure finetune input gradient is allocated at begin and kept until end * remove unnecessary src tensor from ggml_get_rows_back we don't need data of src[2] for computation, only to setup the correct output shape. remove dependency on src[2], so that allocator can work more freely. the computational graph is still completely determined, because the output shape is naturally included. this is similar to how ggml_reshape does it. * remove unnecessary src tensor from ggml_repeat & ggml_repeat_back we don't need data of src[1] for computation, only to setup the correct output shape. remove dependency on src[1], so that allocator can work more freely. the computational graph is still completely determined, because the output shape is naturally included * resolve todo allocator will only make it inplace when they are of the same type * mixing multiple LORA adapters is now possible pass more than one '--lora FNAME' argument to apply more than one LORA. use '--lora-scaled FNAME S' when you want to specify a user-defined scale for an adapter. * add option to save finetune output every N iterations * also save latest finetune output with ITERATION="LATEST" and print where files are saved saving with LATEST makes it easier to resume training from the latest checkpoint the string "LATEST" can be configured with command line option "--fn-latest STR" * update checkpoint train stats before saving via "--save-every" * add command line option `--rank-wo N` for rank of wo tensor * update finetune README * fix dump_non_result_info_yaml to output multiple lora adapters * bug fix: replace GGML_TYPE_SIZE[t] by ggml_type_size(t) * replace llama_n_mult by llama_n_ff * finetune bug fixes to compile with merged in code from master * remove prediction related code to reduce duplicated code with main use main instead * reduce large memory overhead in train-text-from-scratch all gradients had to be pinned so that graph_reset works correctly. this is no longer necessary with the changes to ggml_compute_backward introduced in this PR. * add comment explaining why finetune checkpoints are allocated in one block * make default value of float member a float literal * handle rms_norm and rope parameters the same as in train-text-from-scratch * remove unused code * remove vocab related code as it is unnecessary * add LLM_KV_TRAINING_TYPE to train-text-from-scratch checkpoints so that they can be differentiated from lora finetune checkpoints * add gguf constants and load/save functions from train-text-from-scratch * add load & save lora finetune checkpoints via gguf * add python script to convert old finetune checkpoint files to gguf * remove old checkpoint save & load code * remove code to print data checksums which was used to verify correctness of new gguf code * omit tokenization when training is disabled, only save llama lora adapter training can be disabled by passing '-n 0' to finetune * remove trailing whitespace * update README.md * implement ggml_compute_forward_repeat_f16 * avoid stack overflow of large cgraphs in test-grad0 * add ggml API functions ggml_unravel_index, ggml_get_i32_nd and its analogs for set and for f32 ggml_get_i32_1d, ggml_set_i32_1d, ggml_get_f32_1d, ggml_set_f32_1d now support non-contiguous tensors. in case of non-contiguous tensor, the 1d index is unraveled into a multi index using ggml_unravel_index to be passed to '_nd' function equivalent. this fixes a bug in test-grad0 which happens due to ggml_build_backward not building purely contiguous tensors anymore * increase test-grad0 context mem size to accommodate for bigger cgraph * add sanity check to ggml_compute_backward, asserting the correct shape of gradients * fix ggml_acc_or_set to return tensor of correct shape * remove unused 'inplace' argument from ggml_compute_backward function inplace operations to add gradients are no longer created by ggml_compute_backward use allocator to automatically make inplace operations * add missing argument 'int i0' to ggml_get_i32_nd & ggml_set_i32_nd header declarations * fix error message in ggml_allocr_alloc to display actual max_avail * fix check_gradient ggml_build_backward_expand was previously replaced by ggml_build_backward, but the assignment of forward graph to backward graph missing * use tensor->view_src instead of ggml_is_view and get_view_source * move gradient checkpointing code into ggml, new API function: // build gradient checkpointing backward graph gb for gf using provided checkpoints // gb_tmp will contain original backward graph with rewritten backward process nodes, // but without the second forward pass nodes. GGML_API void ggml_build_backward_gradient_checkpointing( struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, struct ggml_cgraph * gb_tmp, struct ggml_tensor * * checkpoints, int n_checkpoints); * replace custom data getters and setters by ggml functions * train-text-from-scratch can train (full finetune) gguf models just pass the gguf model via `--checkpoint-in FN`. after this, to continue training, pass the generated checkpoint instead of the original gguf model. tested with smaller models, bigger models may exceed available memory. use (LORA) finetune for those. * remove trailing whitespace * add option to save train-text-from-scratch output every N iterations * update README.md * fix warnings * fix warnings * remove finetune option to disable allocator the allocator should always be used. by making sure that it is always used it gets easier to implement automatic memory requirements computation * add tensor checkpoints only when gradient checkpointing is enabled * initialize opt ggml context if none was provided * add ggml-alloc API function 'ggml_allocr_max_size' to get max size of alloc GGML_API size_t ggml_allocr_max_size(struct ggml_allocr * alloc); * finetune: automatically allocate all memory and changes to command line options remove '--n_examples N' parameter, as it no longer makes sense to call optimization process multiple times in a loop. add '--only_write_lora' command line option: will skip tokenization and training, to only write a llama.cpp comptabile LORA adapter. remove memory buffer related command line options. improve iteration console output. * add finetune to Makefile * update README.md * print time per iteration and estimate remaining time * increase measured alloc size by tensor_alignment ggml_allocr_reset will reduce the given size by up to tensor_alignment-1 * fix README.md * add some more allocator debug prints * bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue * revert last commit "bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue" "alloc was freeing an externally allocated tensor, because it calculated the end of allocator memory as alloc->data + alloc->max_size instead of alloc->data + alloc->size." This is intentional to reduce the risk of freeing external tensors when measuring. Unless max_size is not properly calculated, I don't see why this is an issue. * remove unnecessary "0x" before "%p" output * move measurement memory segment to upper region of the address space * update README.md * fix printf format warnings * add missing gguf_free in load_checkpoint_lora_file * load default rms_norm and rope parameters from base model * add gradient accumulation specify number accumulation steps with '--grad-acc N'. this will simulate a bigger batch size of grad_acc*batch. * fix tracking of train_samples and train_tokens * build : fix compile warnings * ggml : fix L-BFGS linesearch loop * improve finetune time measurement fix printf warnings on system where int64_t is (long int). change time datatypes to double because values get big with long training times. exclude file saving from time measurement. converge faster to actual time per iteration by removing very small first duration before first iteration was performed. fix bug in output of total training time, the reported value was 1000 times to small. * specify default lora rank with '--lora-r N' '--lora-r N' will specify default rank for all tensors '--rank-wq N', etc. will override this default rank for specific tensor types. * fix gradient accumulation bug where the same batch was used for each microstep * fix gradient accumulation bug where the same batch was used for each microstep * support grouped-query-attention in ggml_flash_attn and ggml_flash_attn_back k and v can now be repeated in q along ne[2] in forward pass just use modulo to compute k and v indices, like ik2 = iq2 % nek2. in backard pass this won't work as easy, because multiple threads will compete to accumulate to the same k->grad[:,ik1,ik2,ik3] and v->grad[:,iv1,iv2,iv3]. so we change the parallelization over q rows to be over k rows. this ensures non-overlapping (ik2,ik3) across threads. in each thread we then iterate over the number of repetitions of k/v in q to compute iq2 as iq2 = ik2 + irep*nek2. since ne2 is not the same for q,k and v we also change how the gradients are concatenated into the result tensor. additionally the offsets of gradq, gradk and gradv in the result tensor are now memory aligned. we also simplify the compute_backward part of flash_attn to use ggml_reshape instead of switching over the number of dimensions. this needs a small change to ggml_reshape, removing the assertion of second argument to be contiguous. since only the shape (ne) of the second reshape argument is of relevance, its memory layout (nb) is irrelevant -> it can very well be non-contiguous. change test-grad0 to also test for repeated k/v in q. this changes the rng and now results in small gradient differences in softmax. these solely come from using f16 exp table lookup in forward softmax: when temporarily changing softmax to use actual exp function, the reported gradient differences go away. gradient differences coming solely from f16 table lookup are acceptable. added a note to explain this. * add llama API functions to get grouped-query-attention n_head parameter 'n_head_kv'. * fix finetune to support grouped-query-attention (using flash-attention) note: ggml changes to ggml_out_prod are necessary to support grouped-query-attention without flash-attention. * support broadcastable a in out_prod(a, b) and backward pass of broadcasting mul_mat(a, b) * test broadcasting mul_mat backward pass * decouple random number generator of each operation test when changing one test the rng of others tests is not influenced anymore * add comment briefly describing what ggml_repeat_back does * simplify broadcasting mul_mat backward using ggml_repeat_back * add cgraph evaluation order member and corresponding enum type this controls in which order ggml_build_forward visits source nodes. by default the nodes are visited left to right, i.e. src[0] first. in some cases it is beneficial for ggml-alloc to visit in a different order. two possible orders are supported: left-to-right (src[0] first) and right-to-left (src[0] last). * measure max compute size for each cgraph eval order and use best order this can bring huge memory savings: e.g. codellama-34b with n_ctx=64, n_batch=1 goes from 92927.8mb down to 4627.6 MB * remove unused command line options * add sample start patterns and options to force new or by default resume last shuffling * update shuffle rng state on reshuffle * exclude known zero values from computations in flash_attn_f32 & flash_attn_back_f32 * remove probably unnecessary exception type flags from stringstream * pass correct max number of tokens to llama_tokenize * account for possible leading whitespace that will be added by tokenizer e.g. '\t' will be tokenized by llama spm tokenizer to [29871, 12] * use unrolled vec_mad in out_prod y is vec_mad result vec. x is vec_mad input vec. v is vec_mad input scalar. ggml_vec_mad_f32_unroll will internally loop over x and v with same y. GGML_VEC_MAD_UNROLL is by default defined to 32. This value is empirical optimized using performance test runs of out-prod in openllama-3b finetune with 256 context length and batch size 1. It gives 23% performance boost for out_prod. Full measurements of out-prod runtime in ms: unroll_xv unroll_yv 1 67014.643 87826.469 2 77117.552 89077.656 4 72091.311 109121.657 8 61077.543 88678.334 16 56914.67 79514.947 24 59024.595 84350.254 28 55952.446 83368.73 32 51476.658 85177.745 36 55973.792 84659.92 40 55139.616 93844.738 48 60736.392 93330.267 64 99856.878 116994.99 Second column is when unrollying yv instead of xv * set lora_alpha to value of lora_r if it is not set via command line otherwise only changing lora_r will change scaling of lora adapter used in prediction * reshuffle original sample order instead of the previous shuffled order otherwise resumed reshuffle will not result in same sample order * block tiling for out-prod inspired by mul-mat block sizes are empirically optimized roughly doubles the flops of out-prod * exclude some more known zero values from computations in flash_attn_f32 & flash_attn_back_f32 * add static keywords * remove outcommented old code * update train-text-from-scratch with tokenization, sample selection and shuffling from finetune * remove lbfgs related train parameters * move common train functions into common/train.[h|cpp] * move train state into struct train_state * move train data saving code into callback to unify code of opt_callback train_params are still different in finetune and train-text-from-scratch, so it can't yet be moved to train.h|cpp * move common train params into common/train * move common opt_callback into common/train * fix consume_common_train_arg * save and load head_count_kv in lora checkpoints * increase train_samples by used_samples instead of number of batches on batch can contain more than one sample when option "fill_with_next_samples" is used * fix usage of llama_tokenize * remove static from process_escape since we need it exposed in header * fix code formating of long function declarations * fix condition in load_train_state_gguf * use die("msg") instead of replace GGML_ASSERT(!"msg") or throw std::runtime_error("msg") * fix saving and loading of training type * remove terminating '\0' from tokenization (llama_tokenize is now passed the string length instead of relying on terminating '\0') * fix compile warnings * fix compile warnings * use new/delete for train_state instead of malloc/free using malloc may result in seg faults when trying to assign string fields * assert that sample_count > 0, avoiding division by zero * fix frand to return value in interval [0,1) * add train option "--sample-random-offsets" Use samples beginning at random offsets. The offset is only applied to the first sample in each batch context window. Together with "--fill-with-next-samples" this may help for training endless text generation. For example given a dataset containing samples "abcd", "ABCD", "0123". With context size of 8 and options "--fill-with-next-samples", "--no-separate-with-eos", "--no-separate-with-bos", the context windows of batches could only be filled with "abcdABCD", "ABCDabcd", "0123abcd", etc. With "--sample-random-offsets" it can also be filled with "23abcdAB", "bcd0123A", etc. * deduplicate code into function * remove n_rot hparam, as it must always be hparam.n_embd_head() * align code * assert correct base model tensor shapes * move some params from lora hparams into model hparams and load model params from gguf this equalizes the model definition in finetune and text-from-scratch and removes the need for additional llama api functions to get model parameters * remove now unnecessary llama API functions to get model params that where added by this PR * train-text-from-scratch: automatically allocate model tensors, remove option '--mem-model N' * train-text-from-scratch: automatically allocate opt context * train-text-from-scratch: automatically allocate input tensors * train-text-from-scratch: automatically allocate compute memory * remove unused options and equalize train-text-from-scratch with finetune * initialize opt->loss_after with zero * add export-lora program * remove trailing whitespace * add export-lora build in Makefile * remove unused struct tensor_info from export-lora * add export-lora build dependency to llama because it depends on common, which depends on llama * update finetune README.md * cancel optimization when specified number of epochs is completed * improve handling of export-lora arguments print errors and warnings when files could not be read or created * Fix export-lora.cpp "not enough space in the context's memory pool" (#1) * Fix export-lora.cpp "not enough space in the context's memory pool" Without this patch, export-lora would sometimes error with "not enough space in the context's memory pool (needed 656784, available 656800)". * increase required context size by 5*GGML_MEM_ALIGN instead of plain 16 --------- Co-authored-by: xaedes <xaedes@gmail.com> * improve handling of not yet supported tensor types --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: meatbag-18a <145869052+meatbag-18a@users.noreply.github.com>
2023-09-28 18:40:11 +00:00
struct ggml_tensor * llama_get_model_tensor(struct llama_model * model, const char * name) {
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
auto it = std::find_if(model->tensors_by_name.begin(), model->tensors_by_name.end(),
[name](const std::pair<std::string, struct ggml_tensor *> & it) {
return it.first == name;
});
if (it == model->tensors_by_name.end()) {
return nullptr;
}
return it->second;
train : finetune LORA (#2632) * fix track_max_mem in forward_batch_wo_cache_flash_attn_train * remove unnecessary Adam(W) optimizer tensors. reduces optimizer memory overhead from 7*modelsize to 2*modelsize. additionally allows to optimize models with more than 2^31 parameters by replacing int with int64_t. bumps training checkpoint file version, but old checkpoints can still be read. new version with less tensors is saved. * add gradient clipping to AdamW * Fix reset of unused g->nodes and g->grads to NULL * implement gradient checkpointing for training reduces memory overhead from O(n_layer) to O(sqrt(n_layer)) as explained in readme of https://github.com/cybertronai/gradient-checkpointing * remove unused compute buffer 3 * add and use function ggml_build_backward_expand to avoid stack overflows with large maximum number of nodes GGML_API void ggml_build_backward_expand(struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, bool keep); * change AdamW decay parameter to work like the torch AdamW decay parameter It is now relative to Adam learning rate `alpha*sched`. Before that it was relative to `sched` only. `alpha` being the maximum learning rate and `sched` being a scaling parameter in [0..1] * change default AdamW weight decay parameter used in training to 0.1 as used in nanoGPT * change default AdamW weight decay parameter defined in ggml to 0.0, making Adam default instead of AdamW btw: the default weight decay parameter for torch.optim.AdamW is 0.01 * bug fixes for cross entropy loss ggml_cross_entropy_loss: sums where not correctly added in workload of each thread ggml_cross_entropy_loss_back: simplify backward process, reducing numerical issues guard usage of exp f16 lookup in cross entropy by #define GGML_CROSS_ENTROPY_EXP_FP16 cross entropy loss is only used once during training, but it is quite sensitive to numerical errors introduced by exp-f16-lookup. so exp-f16-lookup for cross entropy loss is disabled by default, trading better gradients for very slightly worse runtime performance. * fix test-grad0 for cross_entropy_loss the second argument to cross_entropy_loss must sum up to 1 for each row * fix test-grad0 for soft_max dont use only sum as aggregation, because sum of softmax is always 1 -> finite differences should not work instead use sum(log(soft_max()*(1-eps)+eps)); use eps to avoid log(0) * improve finite differences of test-grad0 by using double instead of float * change cross_entropy_loss to output average over all rows this helps keeping the loss and gradients in a sane range * improve gradient checkpointing sqrt(n_layers) is only the best checkpoint step when mem size of checkpoints and mem size of layers are equal. since layers require more memory than the single-tensor-checkpoint we use, the optimal values are compute different: ``` given: n, u, v objective: minimize(a*u+b*v) where a*b=n, a>0, b>0 b=n/a minimize(a*u+v*n/a) diff(a*u+v*n/a, a) = u - (v*n/a)/a diff(a*u+v*n/a, a) == 0 u - (v*n/a)/a == 0 u == v*n/(a*a) u*a*a = v*n a*a = v*n/u a = sqrt(n*v/u) ``` this change results in more checkpoints, requiring less layers to store between checkpoints, overall improving memory usage. * disable gradient checkpointing debug output * llama : fix rope usage in train-text-from-scratch after ChatGLM change * add more training parameters: --enable-restart N Only for Adam optimizer. Enable restarts of cos-decay --disable-restart N Only for Adam optimizer. Disable restarts of cos-decay --opt-past N Number of optimization iterations to track for delta convergence test. Disabled when zero. --opt-delta N Maximum delta for delta convergence test. Disabled when <= zero. --opt-max-no-improvement N Maximum number of optimization iterations with no improvement. Disabled when <= zero. --adam-epsf N AdamW epsilon for convergence test. Disabled when <= zero. --adam-min-alpha N Adam minimum learning rate alpha, usually 0.1 * alpha * replace memcpy with reshape operation so that the graph is not cut at the input this makes it possible to store other values into the input tensor and then simply recompute the graph without rebuilding it * remove unused function argument from get_example_targets_batch * measure and print total training time * add optimization callback to ggml_opt_resume_g this callback is called before each iteration with custom data and pointer to learning schedule parameter (only used in Adam(W)). can be used for dynamic learning schedule and setting input data for batches before each iteration * use optimization callback in training allows dynamic learning schedule and different batch data for each iteration without relying on low n_iter and high n_examples parameters reduces runtime by avoiding restart of optimization function and improves training convergence by providing a different batch for each iteration * add minimum number of tensor dimensions to apply weight decay (default 2) this allows to not apply weight decay to bias parameters * rename training parameter cos-decay-alpha to cos-decay-min and clarify that adam-min-alpha also applies to warmup * fix increase of model.train_samples and model.train_tokens now that each optimizer iteration gets its own batch we need to multiply by number of opt iterations * change sampling parameters for prediction after training to defaults of common.h and clarify what is context for prediction and what are generated tokens * tighten abs error bounds for cross_entropy_loss in test-grad0 * add conditional compilation of using F16 exp in flash attention uncomment `// #define GGML_FLASH_ATTN_EXP_FP16` to enable usage of f16 exp in flash attention * tighten abs error bounds for flash_attn in test-grad0 * tighten abs error bounds for sqrt in test-grad0 * remove out-commented vectorized code of opt_adam the vectorized code might be bit faster for low number of parameters, but it had a big memory usage overhead * ggml : update ggml_rms_norm_back with configurable eps * llama training : fix ggml_rms_norm_back calls to pass configurable eps * remove trailing whitespace * add train function using automatic gradient checkpointing backward pass and allocator * in train function replace add_inplace by regular add because using add_inplace seems to result in different gradients * don't use allocate hash_map on context because the context has no_alloc=True when using memory allocator resulting in NULL data pointers * correctly clone reshape and permute operations by also cloning tensor->nb values * fix variable name and add missing type cast * terminate recursive tensor cloning when reaching tensor without src tensors * correctly clone view tensors by setting data pointers without this the checkpointing would only work when being used together with memory allocator * fix variable names * swap arguments to commutative ops to be the same as in `forward_batch_wo_cache_flash_attn` * add input tensors as checkpoints so that recursive tensor cloning of gradient checkpointing terminates on input tensors * fix variable name and add missing boolean negation * make sure some tensors are not reallocated by inserting new temporary nodes depending on them: output and parameter gradient tensors need to be available at the end of the graph execution parameter gradient tensors also need to be available before the graph execution because they are set to zero before each optimizer iteration checkpoint tensors are allocated all together to reduce memory allocator fragmentation afterwards, in addition to the temporary nodes, we also need to reset the temporary leafs * fix ASSERT to work with zero layers * add training options whether to use allocator and/or unified training function * integrate unified training function which may use memory allocator the unified training function also supports arguments whether to use flash attention and/or gradient checkpointing * format name of cloned tensors with " (clone)" suffix * set names for tensors in unified train function for easier debugging * allocate graph on context using ggml_new_graph * remove handwritten training functions * remove unused training parameters "use_scratch" and "use_unified" * remove trailing whitespace * remove unused train params: mem_compute1_gb & mem_compute2_gb mem_compute_gb is used for compute when automatic memory allocator is not enabled, otherwise it can be very small to only hold the tensor definitions mem_compute0_gb is used for automatic memory allocator (as long as measurement of max required size is not implemented) * remove unused forward_batch function * add debug asserts in ggml_allocr_alloc to some common pitfalls when using this function directly * only use ggml_allocr_alloc when tensor has NULL data and is no view * fix test when to create temporary backward graph temporary backward graph is only necessary when using checkpointing * fix memory "leak" in optimizers each iteration a new cplan with new memory for work data was allocated. now cplan creation only happens at the start of optimization, with each iteration reusing the cplan and its work data. * reverse order of for loop in ggml_build_backward_expand to save memory when using gradient checkpointing and allocator with this loop order gradient checkpointing with allocator on 16 layer model saves 13% memory; 2 layer memory it saves 2% memory. the computation results are the same * add API functions to access llama model tensors * add stub example for finetuning, based on train-text-from-scratch * move and remove code * add API functions to access remaining model parameters: mult, head and rot * first draft for LORA finetune training * remove const model and layer arguments in API functions for accessing model tensors * bug fixes to make finetune compile automatic allocator does not work yet * add debug prints for training memory improvements * fix names of lora tensors * avoid stack overflow resulting from big ggml_cgraph replace stack allocation and ggml_build_forward by ggml_new_graph in combination with ggml_build_forward_expand * replace llama API functions to get model tensors by one function to get model tensor by name LLAMA_API struct ggml_tensor * llama_get_model_tensor(struct llama_model * model, const char * name); * remove unused call to not existing llama_get_layer_from_model * implement ggml_compute_forward_out_prod_q_f32 * remove trailing whitespace * add lora finetune support on quantized base model tensors * add ggml_add_cast API function this function works like ggml_add, but accepts a data type for the resulting tensor. only supported for quantized src0 input. * use ggml_add_cast in finetuning lora-applied weights will now have data type F32, which improves gradients when finetuning quantized base models * bug fix: actually use result type passed to ggml_add_cast * make sure base model tensors data cannot be used in viewable operations memory allocator would try to make lora application inplace on base model tensors. since those are memory mapped this will result in memory access violations * fix bug in ggml_out_prod which resulted in wrong n_dims of result tensors * avoid keeping in memory ALL of the gradients The problem here stems from ggml_graph_reset. This function is called in the optimization function, before each graph computation, to reset the gradients to zero. This required a unique memory slot for each gradient: allocating memory from a previosly freed memory location might lead to non-zero input gradients. During ggml_compute_backward the gradients are build stepwise by adding or substracting new values, starting from a OP_NONE tensor which needs to contain zero-values. This requires the graph reset. To avoid this I now remember in ggml_build_backward_expand the original OP_NONE gradient tensors in a hash table, which is passed to ggml_compute_backward. There instead of using add (or sub or similar) I test whether the existing gradient to be changed is a zero-valued-tensor by looking up its existence in the hash table. When it is such a zero-tensor it will not be modified, but replaced by the value to be added, otherwise the regular add (not inplace, allocator will take care of this) will be used. This way none of those zero-tensor values will be necessary in the final backward graph and more importantly they won't need a unique memory slot, just to make them zero. * remove trailing whitespace * remove debug prints and function to compute tensor data hash * improve optimization iteration prints * adjust maximal values to support finetuning 3B models * change default finetune params lora_r and lora_alpha to match the n_rank parameters of 4 * bug fix: make sure finetune input gradient is allocated at begin and kept until end * remove unnecessary src tensor from ggml_get_rows_back we don't need data of src[2] for computation, only to setup the correct output shape. remove dependency on src[2], so that allocator can work more freely. the computational graph is still completely determined, because the output shape is naturally included. this is similar to how ggml_reshape does it. * remove unnecessary src tensor from ggml_repeat & ggml_repeat_back we don't need data of src[1] for computation, only to setup the correct output shape. remove dependency on src[1], so that allocator can work more freely. the computational graph is still completely determined, because the output shape is naturally included * resolve todo allocator will only make it inplace when they are of the same type * mixing multiple LORA adapters is now possible pass more than one '--lora FNAME' argument to apply more than one LORA. use '--lora-scaled FNAME S' when you want to specify a user-defined scale for an adapter. * add option to save finetune output every N iterations * also save latest finetune output with ITERATION="LATEST" and print where files are saved saving with LATEST makes it easier to resume training from the latest checkpoint the string "LATEST" can be configured with command line option "--fn-latest STR" * update checkpoint train stats before saving via "--save-every" * add command line option `--rank-wo N` for rank of wo tensor * update finetune README * fix dump_non_result_info_yaml to output multiple lora adapters * bug fix: replace GGML_TYPE_SIZE[t] by ggml_type_size(t) * replace llama_n_mult by llama_n_ff * finetune bug fixes to compile with merged in code from master * remove prediction related code to reduce duplicated code with main use main instead * reduce large memory overhead in train-text-from-scratch all gradients had to be pinned so that graph_reset works correctly. this is no longer necessary with the changes to ggml_compute_backward introduced in this PR. * add comment explaining why finetune checkpoints are allocated in one block * make default value of float member a float literal * handle rms_norm and rope parameters the same as in train-text-from-scratch * remove unused code * remove vocab related code as it is unnecessary * add LLM_KV_TRAINING_TYPE to train-text-from-scratch checkpoints so that they can be differentiated from lora finetune checkpoints * add gguf constants and load/save functions from train-text-from-scratch * add load & save lora finetune checkpoints via gguf * add python script to convert old finetune checkpoint files to gguf * remove old checkpoint save & load code * remove code to print data checksums which was used to verify correctness of new gguf code * omit tokenization when training is disabled, only save llama lora adapter training can be disabled by passing '-n 0' to finetune * remove trailing whitespace * update README.md * implement ggml_compute_forward_repeat_f16 * avoid stack overflow of large cgraphs in test-grad0 * add ggml API functions ggml_unravel_index, ggml_get_i32_nd and its analogs for set and for f32 ggml_get_i32_1d, ggml_set_i32_1d, ggml_get_f32_1d, ggml_set_f32_1d now support non-contiguous tensors. in case of non-contiguous tensor, the 1d index is unraveled into a multi index using ggml_unravel_index to be passed to '_nd' function equivalent. this fixes a bug in test-grad0 which happens due to ggml_build_backward not building purely contiguous tensors anymore * increase test-grad0 context mem size to accommodate for bigger cgraph * add sanity check to ggml_compute_backward, asserting the correct shape of gradients * fix ggml_acc_or_set to return tensor of correct shape * remove unused 'inplace' argument from ggml_compute_backward function inplace operations to add gradients are no longer created by ggml_compute_backward use allocator to automatically make inplace operations * add missing argument 'int i0' to ggml_get_i32_nd & ggml_set_i32_nd header declarations * fix error message in ggml_allocr_alloc to display actual max_avail * fix check_gradient ggml_build_backward_expand was previously replaced by ggml_build_backward, but the assignment of forward graph to backward graph missing * use tensor->view_src instead of ggml_is_view and get_view_source * move gradient checkpointing code into ggml, new API function: // build gradient checkpointing backward graph gb for gf using provided checkpoints // gb_tmp will contain original backward graph with rewritten backward process nodes, // but without the second forward pass nodes. GGML_API void ggml_build_backward_gradient_checkpointing( struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, struct ggml_cgraph * gb_tmp, struct ggml_tensor * * checkpoints, int n_checkpoints); * replace custom data getters and setters by ggml functions * train-text-from-scratch can train (full finetune) gguf models just pass the gguf model via `--checkpoint-in FN`. after this, to continue training, pass the generated checkpoint instead of the original gguf model. tested with smaller models, bigger models may exceed available memory. use (LORA) finetune for those. * remove trailing whitespace * add option to save train-text-from-scratch output every N iterations * update README.md * fix warnings * fix warnings * remove finetune option to disable allocator the allocator should always be used. by making sure that it is always used it gets easier to implement automatic memory requirements computation * add tensor checkpoints only when gradient checkpointing is enabled * initialize opt ggml context if none was provided * add ggml-alloc API function 'ggml_allocr_max_size' to get max size of alloc GGML_API size_t ggml_allocr_max_size(struct ggml_allocr * alloc); * finetune: automatically allocate all memory and changes to command line options remove '--n_examples N' parameter, as it no longer makes sense to call optimization process multiple times in a loop. add '--only_write_lora' command line option: will skip tokenization and training, to only write a llama.cpp comptabile LORA adapter. remove memory buffer related command line options. improve iteration console output. * add finetune to Makefile * update README.md * print time per iteration and estimate remaining time * increase measured alloc size by tensor_alignment ggml_allocr_reset will reduce the given size by up to tensor_alignment-1 * fix README.md * add some more allocator debug prints * bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue * revert last commit "bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue" "alloc was freeing an externally allocated tensor, because it calculated the end of allocator memory as alloc->data + alloc->max_size instead of alloc->data + alloc->size." This is intentional to reduce the risk of freeing external tensors when measuring. Unless max_size is not properly calculated, I don't see why this is an issue. * remove unnecessary "0x" before "%p" output * move measurement memory segment to upper region of the address space * update README.md * fix printf format warnings * add missing gguf_free in load_checkpoint_lora_file * load default rms_norm and rope parameters from base model * add gradient accumulation specify number accumulation steps with '--grad-acc N'. this will simulate a bigger batch size of grad_acc*batch. * fix tracking of train_samples and train_tokens * build : fix compile warnings * ggml : fix L-BFGS linesearch loop * improve finetune time measurement fix printf warnings on system where int64_t is (long int). change time datatypes to double because values get big with long training times. exclude file saving from time measurement. converge faster to actual time per iteration by removing very small first duration before first iteration was performed. fix bug in output of total training time, the reported value was 1000 times to small. * specify default lora rank with '--lora-r N' '--lora-r N' will specify default rank for all tensors '--rank-wq N', etc. will override this default rank for specific tensor types. * fix gradient accumulation bug where the same batch was used for each microstep * fix gradient accumulation bug where the same batch was used for each microstep * support grouped-query-attention in ggml_flash_attn and ggml_flash_attn_back k and v can now be repeated in q along ne[2] in forward pass just use modulo to compute k and v indices, like ik2 = iq2 % nek2. in backard pass this won't work as easy, because multiple threads will compete to accumulate to the same k->grad[:,ik1,ik2,ik3] and v->grad[:,iv1,iv2,iv3]. so we change the parallelization over q rows to be over k rows. this ensures non-overlapping (ik2,ik3) across threads. in each thread we then iterate over the number of repetitions of k/v in q to compute iq2 as iq2 = ik2 + irep*nek2. since ne2 is not the same for q,k and v we also change how the gradients are concatenated into the result tensor. additionally the offsets of gradq, gradk and gradv in the result tensor are now memory aligned. we also simplify the compute_backward part of flash_attn to use ggml_reshape instead of switching over the number of dimensions. this needs a small change to ggml_reshape, removing the assertion of second argument to be contiguous. since only the shape (ne) of the second reshape argument is of relevance, its memory layout (nb) is irrelevant -> it can very well be non-contiguous. change test-grad0 to also test for repeated k/v in q. this changes the rng and now results in small gradient differences in softmax. these solely come from using f16 exp table lookup in forward softmax: when temporarily changing softmax to use actual exp function, the reported gradient differences go away. gradient differences coming solely from f16 table lookup are acceptable. added a note to explain this. * add llama API functions to get grouped-query-attention n_head parameter 'n_head_kv'. * fix finetune to support grouped-query-attention (using flash-attention) note: ggml changes to ggml_out_prod are necessary to support grouped-query-attention without flash-attention. * support broadcastable a in out_prod(a, b) and backward pass of broadcasting mul_mat(a, b) * test broadcasting mul_mat backward pass * decouple random number generator of each operation test when changing one test the rng of others tests is not influenced anymore * add comment briefly describing what ggml_repeat_back does * simplify broadcasting mul_mat backward using ggml_repeat_back * add cgraph evaluation order member and corresponding enum type this controls in which order ggml_build_forward visits source nodes. by default the nodes are visited left to right, i.e. src[0] first. in some cases it is beneficial for ggml-alloc to visit in a different order. two possible orders are supported: left-to-right (src[0] first) and right-to-left (src[0] last). * measure max compute size for each cgraph eval order and use best order this can bring huge memory savings: e.g. codellama-34b with n_ctx=64, n_batch=1 goes from 92927.8mb down to 4627.6 MB * remove unused command line options * add sample start patterns and options to force new or by default resume last shuffling * update shuffle rng state on reshuffle * exclude known zero values from computations in flash_attn_f32 & flash_attn_back_f32 * remove probably unnecessary exception type flags from stringstream * pass correct max number of tokens to llama_tokenize * account for possible leading whitespace that will be added by tokenizer e.g. '\t' will be tokenized by llama spm tokenizer to [29871, 12] * use unrolled vec_mad in out_prod y is vec_mad result vec. x is vec_mad input vec. v is vec_mad input scalar. ggml_vec_mad_f32_unroll will internally loop over x and v with same y. GGML_VEC_MAD_UNROLL is by default defined to 32. This value is empirical optimized using performance test runs of out-prod in openllama-3b finetune with 256 context length and batch size 1. It gives 23% performance boost for out_prod. Full measurements of out-prod runtime in ms: unroll_xv unroll_yv 1 67014.643 87826.469 2 77117.552 89077.656 4 72091.311 109121.657 8 61077.543 88678.334 16 56914.67 79514.947 24 59024.595 84350.254 28 55952.446 83368.73 32 51476.658 85177.745 36 55973.792 84659.92 40 55139.616 93844.738 48 60736.392 93330.267 64 99856.878 116994.99 Second column is when unrollying yv instead of xv * set lora_alpha to value of lora_r if it is not set via command line otherwise only changing lora_r will change scaling of lora adapter used in prediction * reshuffle original sample order instead of the previous shuffled order otherwise resumed reshuffle will not result in same sample order * block tiling for out-prod inspired by mul-mat block sizes are empirically optimized roughly doubles the flops of out-prod * exclude some more known zero values from computations in flash_attn_f32 & flash_attn_back_f32 * add static keywords * remove outcommented old code * update train-text-from-scratch with tokenization, sample selection and shuffling from finetune * remove lbfgs related train parameters * move common train functions into common/train.[h|cpp] * move train state into struct train_state * move train data saving code into callback to unify code of opt_callback train_params are still different in finetune and train-text-from-scratch, so it can't yet be moved to train.h|cpp * move common train params into common/train * move common opt_callback into common/train * fix consume_common_train_arg * save and load head_count_kv in lora checkpoints * increase train_samples by used_samples instead of number of batches on batch can contain more than one sample when option "fill_with_next_samples" is used * fix usage of llama_tokenize * remove static from process_escape since we need it exposed in header * fix code formating of long function declarations * fix condition in load_train_state_gguf * use die("msg") instead of replace GGML_ASSERT(!"msg") or throw std::runtime_error("msg") * fix saving and loading of training type * remove terminating '\0' from tokenization (llama_tokenize is now passed the string length instead of relying on terminating '\0') * fix compile warnings * fix compile warnings * use new/delete for train_state instead of malloc/free using malloc may result in seg faults when trying to assign string fields * assert that sample_count > 0, avoiding division by zero * fix frand to return value in interval [0,1) * add train option "--sample-random-offsets" Use samples beginning at random offsets. The offset is only applied to the first sample in each batch context window. Together with "--fill-with-next-samples" this may help for training endless text generation. For example given a dataset containing samples "abcd", "ABCD", "0123". With context size of 8 and options "--fill-with-next-samples", "--no-separate-with-eos", "--no-separate-with-bos", the context windows of batches could only be filled with "abcdABCD", "ABCDabcd", "0123abcd", etc. With "--sample-random-offsets" it can also be filled with "23abcdAB", "bcd0123A", etc. * deduplicate code into function * remove n_rot hparam, as it must always be hparam.n_embd_head() * align code * assert correct base model tensor shapes * move some params from lora hparams into model hparams and load model params from gguf this equalizes the model definition in finetune and text-from-scratch and removes the need for additional llama api functions to get model parameters * remove now unnecessary llama API functions to get model params that where added by this PR * train-text-from-scratch: automatically allocate model tensors, remove option '--mem-model N' * train-text-from-scratch: automatically allocate opt context * train-text-from-scratch: automatically allocate input tensors * train-text-from-scratch: automatically allocate compute memory * remove unused options and equalize train-text-from-scratch with finetune * initialize opt->loss_after with zero * add export-lora program * remove trailing whitespace * add export-lora build in Makefile * remove unused struct tensor_info from export-lora * add export-lora build dependency to llama because it depends on common, which depends on llama * update finetune README.md * cancel optimization when specified number of epochs is completed * improve handling of export-lora arguments print errors and warnings when files could not be read or created * Fix export-lora.cpp "not enough space in the context's memory pool" (#1) * Fix export-lora.cpp "not enough space in the context's memory pool" Without this patch, export-lora would sometimes error with "not enough space in the context's memory pool (needed 656784, available 656800)". * increase required context size by 5*GGML_MEM_ALIGN instead of plain 16 --------- Co-authored-by: xaedes <xaedes@gmail.com> * improve handling of not yet supported tensor types --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: meatbag-18a <145869052+meatbag-18a@users.noreply.github.com>
2023-09-28 18:40:11 +00:00
}
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
uint32_t llama_model_quantize(
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
const char * fname_inp,
const char * fname_out,
const llama_model_quantize_params * params) {
try {
llama_model_quantize_internal(fname_inp, fname_out, params);
return 0;
} catch (const std::exception & err) {
LLAMA_LOG_ERROR("%s: failed to quantize: %s\n", __func__, err.what());
return 1;
}
Add LoRA support (#820)
2023-04-17 15:28:55 +00:00
}
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
int32_t llama_model_apply_lora_from_file(const struct llama_model * model, const char * path_lora, float scale, const char * path_base_model, int32_t n_threads) {
llama : make model stateless and context stateful (llama_state) (#1797) * llama : make model stateless and context stateful * llama : minor cleanup * llama : update internal API declaration * Apply suggestions from code review fix style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Missing model memory release * Fix style * Add deprecated warning for public API function llama_init_from_file * Update public API use cases: move away from deprecated llama_init_from_file * Deprecate public API function llama_apply_lora_from_file --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-24 08:47:58 +00:00
try {
train : finetune LORA (#2632) * fix track_max_mem in forward_batch_wo_cache_flash_attn_train * remove unnecessary Adam(W) optimizer tensors. reduces optimizer memory overhead from 7*modelsize to 2*modelsize. additionally allows to optimize models with more than 2^31 parameters by replacing int with int64_t. bumps training checkpoint file version, but old checkpoints can still be read. new version with less tensors is saved. * add gradient clipping to AdamW * Fix reset of unused g->nodes and g->grads to NULL * implement gradient checkpointing for training reduces memory overhead from O(n_layer) to O(sqrt(n_layer)) as explained in readme of https://github.com/cybertronai/gradient-checkpointing * remove unused compute buffer 3 * add and use function ggml_build_backward_expand to avoid stack overflows with large maximum number of nodes GGML_API void ggml_build_backward_expand(struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, bool keep); * change AdamW decay parameter to work like the torch AdamW decay parameter It is now relative to Adam learning rate `alpha*sched`. Before that it was relative to `sched` only. `alpha` being the maximum learning rate and `sched` being a scaling parameter in [0..1] * change default AdamW weight decay parameter used in training to 0.1 as used in nanoGPT * change default AdamW weight decay parameter defined in ggml to 0.0, making Adam default instead of AdamW btw: the default weight decay parameter for torch.optim.AdamW is 0.01 * bug fixes for cross entropy loss ggml_cross_entropy_loss: sums where not correctly added in workload of each thread ggml_cross_entropy_loss_back: simplify backward process, reducing numerical issues guard usage of exp f16 lookup in cross entropy by #define GGML_CROSS_ENTROPY_EXP_FP16 cross entropy loss is only used once during training, but it is quite sensitive to numerical errors introduced by exp-f16-lookup. so exp-f16-lookup for cross entropy loss is disabled by default, trading better gradients for very slightly worse runtime performance. * fix test-grad0 for cross_entropy_loss the second argument to cross_entropy_loss must sum up to 1 for each row * fix test-grad0 for soft_max dont use only sum as aggregation, because sum of softmax is always 1 -> finite differences should not work instead use sum(log(soft_max()*(1-eps)+eps)); use eps to avoid log(0) * improve finite differences of test-grad0 by using double instead of float * change cross_entropy_loss to output average over all rows this helps keeping the loss and gradients in a sane range * improve gradient checkpointing sqrt(n_layers) is only the best checkpoint step when mem size of checkpoints and mem size of layers are equal. since layers require more memory than the single-tensor-checkpoint we use, the optimal values are compute different: ``` given: n, u, v objective: minimize(a*u+b*v) where a*b=n, a>0, b>0 b=n/a minimize(a*u+v*n/a) diff(a*u+v*n/a, a) = u - (v*n/a)/a diff(a*u+v*n/a, a) == 0 u - (v*n/a)/a == 0 u == v*n/(a*a) u*a*a = v*n a*a = v*n/u a = sqrt(n*v/u) ``` this change results in more checkpoints, requiring less layers to store between checkpoints, overall improving memory usage. * disable gradient checkpointing debug output * llama : fix rope usage in train-text-from-scratch after ChatGLM change * add more training parameters: --enable-restart N Only for Adam optimizer. Enable restarts of cos-decay --disable-restart N Only for Adam optimizer. Disable restarts of cos-decay --opt-past N Number of optimization iterations to track for delta convergence test. Disabled when zero. --opt-delta N Maximum delta for delta convergence test. Disabled when <= zero. --opt-max-no-improvement N Maximum number of optimization iterations with no improvement. Disabled when <= zero. --adam-epsf N AdamW epsilon for convergence test. Disabled when <= zero. --adam-min-alpha N Adam minimum learning rate alpha, usually 0.1 * alpha * replace memcpy with reshape operation so that the graph is not cut at the input this makes it possible to store other values into the input tensor and then simply recompute the graph without rebuilding it * remove unused function argument from get_example_targets_batch * measure and print total training time * add optimization callback to ggml_opt_resume_g this callback is called before each iteration with custom data and pointer to learning schedule parameter (only used in Adam(W)). can be used for dynamic learning schedule and setting input data for batches before each iteration * use optimization callback in training allows dynamic learning schedule and different batch data for each iteration without relying on low n_iter and high n_examples parameters reduces runtime by avoiding restart of optimization function and improves training convergence by providing a different batch for each iteration * add minimum number of tensor dimensions to apply weight decay (default 2) this allows to not apply weight decay to bias parameters * rename training parameter cos-decay-alpha to cos-decay-min and clarify that adam-min-alpha also applies to warmup * fix increase of model.train_samples and model.train_tokens now that each optimizer iteration gets its own batch we need to multiply by number of opt iterations * change sampling parameters for prediction after training to defaults of common.h and clarify what is context for prediction and what are generated tokens * tighten abs error bounds for cross_entropy_loss in test-grad0 * add conditional compilation of using F16 exp in flash attention uncomment `// #define GGML_FLASH_ATTN_EXP_FP16` to enable usage of f16 exp in flash attention * tighten abs error bounds for flash_attn in test-grad0 * tighten abs error bounds for sqrt in test-grad0 * remove out-commented vectorized code of opt_adam the vectorized code might be bit faster for low number of parameters, but it had a big memory usage overhead * ggml : update ggml_rms_norm_back with configurable eps * llama training : fix ggml_rms_norm_back calls to pass configurable eps * remove trailing whitespace * add train function using automatic gradient checkpointing backward pass and allocator * in train function replace add_inplace by regular add because using add_inplace seems to result in different gradients * don't use allocate hash_map on context because the context has no_alloc=True when using memory allocator resulting in NULL data pointers * correctly clone reshape and permute operations by also cloning tensor->nb values * fix variable name and add missing type cast * terminate recursive tensor cloning when reaching tensor without src tensors * correctly clone view tensors by setting data pointers without this the checkpointing would only work when being used together with memory allocator * fix variable names * swap arguments to commutative ops to be the same as in `forward_batch_wo_cache_flash_attn` * add input tensors as checkpoints so that recursive tensor cloning of gradient checkpointing terminates on input tensors * fix variable name and add missing boolean negation * make sure some tensors are not reallocated by inserting new temporary nodes depending on them: output and parameter gradient tensors need to be available at the end of the graph execution parameter gradient tensors also need to be available before the graph execution because they are set to zero before each optimizer iteration checkpoint tensors are allocated all together to reduce memory allocator fragmentation afterwards, in addition to the temporary nodes, we also need to reset the temporary leafs * fix ASSERT to work with zero layers * add training options whether to use allocator and/or unified training function * integrate unified training function which may use memory allocator the unified training function also supports arguments whether to use flash attention and/or gradient checkpointing * format name of cloned tensors with " (clone)" suffix * set names for tensors in unified train function for easier debugging * allocate graph on context using ggml_new_graph * remove handwritten training functions * remove unused training parameters "use_scratch" and "use_unified" * remove trailing whitespace * remove unused train params: mem_compute1_gb & mem_compute2_gb mem_compute_gb is used for compute when automatic memory allocator is not enabled, otherwise it can be very small to only hold the tensor definitions mem_compute0_gb is used for automatic memory allocator (as long as measurement of max required size is not implemented) * remove unused forward_batch function * add debug asserts in ggml_allocr_alloc to some common pitfalls when using this function directly * only use ggml_allocr_alloc when tensor has NULL data and is no view * fix test when to create temporary backward graph temporary backward graph is only necessary when using checkpointing * fix memory "leak" in optimizers each iteration a new cplan with new memory for work data was allocated. now cplan creation only happens at the start of optimization, with each iteration reusing the cplan and its work data. * reverse order of for loop in ggml_build_backward_expand to save memory when using gradient checkpointing and allocator with this loop order gradient checkpointing with allocator on 16 layer model saves 13% memory; 2 layer memory it saves 2% memory. the computation results are the same * add API functions to access llama model tensors * add stub example for finetuning, based on train-text-from-scratch * move and remove code * add API functions to access remaining model parameters: mult, head and rot * first draft for LORA finetune training * remove const model and layer arguments in API functions for accessing model tensors * bug fixes to make finetune compile automatic allocator does not work yet * add debug prints for training memory improvements * fix names of lora tensors * avoid stack overflow resulting from big ggml_cgraph replace stack allocation and ggml_build_forward by ggml_new_graph in combination with ggml_build_forward_expand * replace llama API functions to get model tensors by one function to get model tensor by name LLAMA_API struct ggml_tensor * llama_get_model_tensor(struct llama_model * model, const char * name); * remove unused call to not existing llama_get_layer_from_model * implement ggml_compute_forward_out_prod_q_f32 * remove trailing whitespace * add lora finetune support on quantized base model tensors * add ggml_add_cast API function this function works like ggml_add, but accepts a data type for the resulting tensor. only supported for quantized src0 input. * use ggml_add_cast in finetuning lora-applied weights will now have data type F32, which improves gradients when finetuning quantized base models * bug fix: actually use result type passed to ggml_add_cast * make sure base model tensors data cannot be used in viewable operations memory allocator would try to make lora application inplace on base model tensors. since those are memory mapped this will result in memory access violations * fix bug in ggml_out_prod which resulted in wrong n_dims of result tensors * avoid keeping in memory ALL of the gradients The problem here stems from ggml_graph_reset. This function is called in the optimization function, before each graph computation, to reset the gradients to zero. This required a unique memory slot for each gradient: allocating memory from a previosly freed memory location might lead to non-zero input gradients. During ggml_compute_backward the gradients are build stepwise by adding or substracting new values, starting from a OP_NONE tensor which needs to contain zero-values. This requires the graph reset. To avoid this I now remember in ggml_build_backward_expand the original OP_NONE gradient tensors in a hash table, which is passed to ggml_compute_backward. There instead of using add (or sub or similar) I test whether the existing gradient to be changed is a zero-valued-tensor by looking up its existence in the hash table. When it is such a zero-tensor it will not be modified, but replaced by the value to be added, otherwise the regular add (not inplace, allocator will take care of this) will be used. This way none of those zero-tensor values will be necessary in the final backward graph and more importantly they won't need a unique memory slot, just to make them zero. * remove trailing whitespace * remove debug prints and function to compute tensor data hash * improve optimization iteration prints * adjust maximal values to support finetuning 3B models * change default finetune params lora_r and lora_alpha to match the n_rank parameters of 4 * bug fix: make sure finetune input gradient is allocated at begin and kept until end * remove unnecessary src tensor from ggml_get_rows_back we don't need data of src[2] for computation, only to setup the correct output shape. remove dependency on src[2], so that allocator can work more freely. the computational graph is still completely determined, because the output shape is naturally included. this is similar to how ggml_reshape does it. * remove unnecessary src tensor from ggml_repeat & ggml_repeat_back we don't need data of src[1] for computation, only to setup the correct output shape. remove dependency on src[1], so that allocator can work more freely. the computational graph is still completely determined, because the output shape is naturally included * resolve todo allocator will only make it inplace when they are of the same type * mixing multiple LORA adapters is now possible pass more than one '--lora FNAME' argument to apply more than one LORA. use '--lora-scaled FNAME S' when you want to specify a user-defined scale for an adapter. * add option to save finetune output every N iterations * also save latest finetune output with ITERATION="LATEST" and print where files are saved saving with LATEST makes it easier to resume training from the latest checkpoint the string "LATEST" can be configured with command line option "--fn-latest STR" * update checkpoint train stats before saving via "--save-every" * add command line option `--rank-wo N` for rank of wo tensor * update finetune README * fix dump_non_result_info_yaml to output multiple lora adapters * bug fix: replace GGML_TYPE_SIZE[t] by ggml_type_size(t) * replace llama_n_mult by llama_n_ff * finetune bug fixes to compile with merged in code from master * remove prediction related code to reduce duplicated code with main use main instead * reduce large memory overhead in train-text-from-scratch all gradients had to be pinned so that graph_reset works correctly. this is no longer necessary with the changes to ggml_compute_backward introduced in this PR. * add comment explaining why finetune checkpoints are allocated in one block * make default value of float member a float literal * handle rms_norm and rope parameters the same as in train-text-from-scratch * remove unused code * remove vocab related code as it is unnecessary * add LLM_KV_TRAINING_TYPE to train-text-from-scratch checkpoints so that they can be differentiated from lora finetune checkpoints * add gguf constants and load/save functions from train-text-from-scratch * add load & save lora finetune checkpoints via gguf * add python script to convert old finetune checkpoint files to gguf * remove old checkpoint save & load code * remove code to print data checksums which was used to verify correctness of new gguf code * omit tokenization when training is disabled, only save llama lora adapter training can be disabled by passing '-n 0' to finetune * remove trailing whitespace * update README.md * implement ggml_compute_forward_repeat_f16 * avoid stack overflow of large cgraphs in test-grad0 * add ggml API functions ggml_unravel_index, ggml_get_i32_nd and its analogs for set and for f32 ggml_get_i32_1d, ggml_set_i32_1d, ggml_get_f32_1d, ggml_set_f32_1d now support non-contiguous tensors. in case of non-contiguous tensor, the 1d index is unraveled into a multi index using ggml_unravel_index to be passed to '_nd' function equivalent. this fixes a bug in test-grad0 which happens due to ggml_build_backward not building purely contiguous tensors anymore * increase test-grad0 context mem size to accommodate for bigger cgraph * add sanity check to ggml_compute_backward, asserting the correct shape of gradients * fix ggml_acc_or_set to return tensor of correct shape * remove unused 'inplace' argument from ggml_compute_backward function inplace operations to add gradients are no longer created by ggml_compute_backward use allocator to automatically make inplace operations * add missing argument 'int i0' to ggml_get_i32_nd & ggml_set_i32_nd header declarations * fix error message in ggml_allocr_alloc to display actual max_avail * fix check_gradient ggml_build_backward_expand was previously replaced by ggml_build_backward, but the assignment of forward graph to backward graph missing * use tensor->view_src instead of ggml_is_view and get_view_source * move gradient checkpointing code into ggml, new API function: // build gradient checkpointing backward graph gb for gf using provided checkpoints // gb_tmp will contain original backward graph with rewritten backward process nodes, // but without the second forward pass nodes. GGML_API void ggml_build_backward_gradient_checkpointing( struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, struct ggml_cgraph * gb_tmp, struct ggml_tensor * * checkpoints, int n_checkpoints); * replace custom data getters and setters by ggml functions * train-text-from-scratch can train (full finetune) gguf models just pass the gguf model via `--checkpoint-in FN`. after this, to continue training, pass the generated checkpoint instead of the original gguf model. tested with smaller models, bigger models may exceed available memory. use (LORA) finetune for those. * remove trailing whitespace * add option to save train-text-from-scratch output every N iterations * update README.md * fix warnings * fix warnings * remove finetune option to disable allocator the allocator should always be used. by making sure that it is always used it gets easier to implement automatic memory requirements computation * add tensor checkpoints only when gradient checkpointing is enabled * initialize opt ggml context if none was provided * add ggml-alloc API function 'ggml_allocr_max_size' to get max size of alloc GGML_API size_t ggml_allocr_max_size(struct ggml_allocr * alloc); * finetune: automatically allocate all memory and changes to command line options remove '--n_examples N' parameter, as it no longer makes sense to call optimization process multiple times in a loop. add '--only_write_lora' command line option: will skip tokenization and training, to only write a llama.cpp comptabile LORA adapter. remove memory buffer related command line options. improve iteration console output. * add finetune to Makefile * update README.md * print time per iteration and estimate remaining time * increase measured alloc size by tensor_alignment ggml_allocr_reset will reduce the given size by up to tensor_alignment-1 * fix README.md * add some more allocator debug prints * bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue * revert last commit "bug fix, probably solves the 'ggml_allocr_alloc: not enough space in the buffer' issue" "alloc was freeing an externally allocated tensor, because it calculated the end of allocator memory as alloc->data + alloc->max_size instead of alloc->data + alloc->size." This is intentional to reduce the risk of freeing external tensors when measuring. Unless max_size is not properly calculated, I don't see why this is an issue. * remove unnecessary "0x" before "%p" output * move measurement memory segment to upper region of the address space * update README.md * fix printf format warnings * add missing gguf_free in load_checkpoint_lora_file * load default rms_norm and rope parameters from base model * add gradient accumulation specify number accumulation steps with '--grad-acc N'. this will simulate a bigger batch size of grad_acc*batch. * fix tracking of train_samples and train_tokens * build : fix compile warnings * ggml : fix L-BFGS linesearch loop * improve finetune time measurement fix printf warnings on system where int64_t is (long int). change time datatypes to double because values get big with long training times. exclude file saving from time measurement. converge faster to actual time per iteration by removing very small first duration before first iteration was performed. fix bug in output of total training time, the reported value was 1000 times to small. * specify default lora rank with '--lora-r N' '--lora-r N' will specify default rank for all tensors '--rank-wq N', etc. will override this default rank for specific tensor types. * fix gradient accumulation bug where the same batch was used for each microstep * fix gradient accumulation bug where the same batch was used for each microstep * support grouped-query-attention in ggml_flash_attn and ggml_flash_attn_back k and v can now be repeated in q along ne[2] in forward pass just use modulo to compute k and v indices, like ik2 = iq2 % nek2. in backard pass this won't work as easy, because multiple threads will compete to accumulate to the same k->grad[:,ik1,ik2,ik3] and v->grad[:,iv1,iv2,iv3]. so we change the parallelization over q rows to be over k rows. this ensures non-overlapping (ik2,ik3) across threads. in each thread we then iterate over the number of repetitions of k/v in q to compute iq2 as iq2 = ik2 + irep*nek2. since ne2 is not the same for q,k and v we also change how the gradients are concatenated into the result tensor. additionally the offsets of gradq, gradk and gradv in the result tensor are now memory aligned. we also simplify the compute_backward part of flash_attn to use ggml_reshape instead of switching over the number of dimensions. this needs a small change to ggml_reshape, removing the assertion of second argument to be contiguous. since only the shape (ne) of the second reshape argument is of relevance, its memory layout (nb) is irrelevant -> it can very well be non-contiguous. change test-grad0 to also test for repeated k/v in q. this changes the rng and now results in small gradient differences in softmax. these solely come from using f16 exp table lookup in forward softmax: when temporarily changing softmax to use actual exp function, the reported gradient differences go away. gradient differences coming solely from f16 table lookup are acceptable. added a note to explain this. * add llama API functions to get grouped-query-attention n_head parameter 'n_head_kv'. * fix finetune to support grouped-query-attention (using flash-attention) note: ggml changes to ggml_out_prod are necessary to support grouped-query-attention without flash-attention. * support broadcastable a in out_prod(a, b) and backward pass of broadcasting mul_mat(a, b) * test broadcasting mul_mat backward pass * decouple random number generator of each operation test when changing one test the rng of others tests is not influenced anymore * add comment briefly describing what ggml_repeat_back does * simplify broadcasting mul_mat backward using ggml_repeat_back * add cgraph evaluation order member and corresponding enum type this controls in which order ggml_build_forward visits source nodes. by default the nodes are visited left to right, i.e. src[0] first. in some cases it is beneficial for ggml-alloc to visit in a different order. two possible orders are supported: left-to-right (src[0] first) and right-to-left (src[0] last). * measure max compute size for each cgraph eval order and use best order this can bring huge memory savings: e.g. codellama-34b with n_ctx=64, n_batch=1 goes from 92927.8mb down to 4627.6 MB * remove unused command line options * add sample start patterns and options to force new or by default resume last shuffling * update shuffle rng state on reshuffle * exclude known zero values from computations in flash_attn_f32 & flash_attn_back_f32 * remove probably unnecessary exception type flags from stringstream * pass correct max number of tokens to llama_tokenize * account for possible leading whitespace that will be added by tokenizer e.g. '\t' will be tokenized by llama spm tokenizer to [29871, 12] * use unrolled vec_mad in out_prod y is vec_mad result vec. x is vec_mad input vec. v is vec_mad input scalar. ggml_vec_mad_f32_unroll will internally loop over x and v with same y. GGML_VEC_MAD_UNROLL is by default defined to 32. This value is empirical optimized using performance test runs of out-prod in openllama-3b finetune with 256 context length and batch size 1. It gives 23% performance boost for out_prod. Full measurements of out-prod runtime in ms: unroll_xv unroll_yv 1 67014.643 87826.469 2 77117.552 89077.656 4 72091.311 109121.657 8 61077.543 88678.334 16 56914.67 79514.947 24 59024.595 84350.254 28 55952.446 83368.73 32 51476.658 85177.745 36 55973.792 84659.92 40 55139.616 93844.738 48 60736.392 93330.267 64 99856.878 116994.99 Second column is when unrollying yv instead of xv * set lora_alpha to value of lora_r if it is not set via command line otherwise only changing lora_r will change scaling of lora adapter used in prediction * reshuffle original sample order instead of the previous shuffled order otherwise resumed reshuffle will not result in same sample order * block tiling for out-prod inspired by mul-mat block sizes are empirically optimized roughly doubles the flops of out-prod * exclude some more known zero values from computations in flash_attn_f32 & flash_attn_back_f32 * add static keywords * remove outcommented old code * update train-text-from-scratch with tokenization, sample selection and shuffling from finetune * remove lbfgs related train parameters * move common train functions into common/train.[h|cpp] * move train state into struct train_state * move train data saving code into callback to unify code of opt_callback train_params are still different in finetune and train-text-from-scratch, so it can't yet be moved to train.h|cpp * move common train params into common/train * move common opt_callback into common/train * fix consume_common_train_arg * save and load head_count_kv in lora checkpoints * increase train_samples by used_samples instead of number of batches on batch can contain more than one sample when option "fill_with_next_samples" is used * fix usage of llama_tokenize * remove static from process_escape since we need it exposed in header * fix code formating of long function declarations * fix condition in load_train_state_gguf * use die("msg") instead of replace GGML_ASSERT(!"msg") or throw std::runtime_error("msg") * fix saving and loading of training type * remove terminating '\0' from tokenization (llama_tokenize is now passed the string length instead of relying on terminating '\0') * fix compile warnings * fix compile warnings * use new/delete for train_state instead of malloc/free using malloc may result in seg faults when trying to assign string fields * assert that sample_count > 0, avoiding division by zero * fix frand to return value in interval [0,1) * add train option "--sample-random-offsets" Use samples beginning at random offsets. The offset is only applied to the first sample in each batch context window. Together with "--fill-with-next-samples" this may help for training endless text generation. For example given a dataset containing samples "abcd", "ABCD", "0123". With context size of 8 and options "--fill-with-next-samples", "--no-separate-with-eos", "--no-separate-with-bos", the context windows of batches could only be filled with "abcdABCD", "ABCDabcd", "0123abcd", etc. With "--sample-random-offsets" it can also be filled with "23abcdAB", "bcd0123A", etc. * deduplicate code into function * remove n_rot hparam, as it must always be hparam.n_embd_head() * align code * assert correct base model tensor shapes * move some params from lora hparams into model hparams and load model params from gguf this equalizes the model definition in finetune and text-from-scratch and removes the need for additional llama api functions to get model parameters * remove now unnecessary llama API functions to get model params that where added by this PR * train-text-from-scratch: automatically allocate model tensors, remove option '--mem-model N' * train-text-from-scratch: automatically allocate opt context * train-text-from-scratch: automatically allocate input tensors * train-text-from-scratch: automatically allocate compute memory * remove unused options and equalize train-text-from-scratch with finetune * initialize opt->loss_after with zero * add export-lora program * remove trailing whitespace * add export-lora build in Makefile * remove unused struct tensor_info from export-lora * add export-lora build dependency to llama because it depends on common, which depends on llama * update finetune README.md * cancel optimization when specified number of epochs is completed * improve handling of export-lora arguments print errors and warnings when files could not be read or created * Fix export-lora.cpp "not enough space in the context's memory pool" (#1) * Fix export-lora.cpp "not enough space in the context's memory pool" Without this patch, export-lora would sometimes error with "not enough space in the context's memory pool (needed 656784, available 656800)". * increase required context size by 5*GGML_MEM_ALIGN instead of plain 16 --------- Co-authored-by: xaedes <xaedes@gmail.com> * improve handling of not yet supported tensor types --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: meatbag-18a <145869052+meatbag-18a@users.noreply.github.com>
2023-09-28 18:40:11 +00:00
return llama_apply_lora_from_file_internal(*model, path_lora, scale, path_base_model, n_threads);
llama : consistently catch and throw only exceptions deriving from std::exception (#1599) Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-05 20:24:29 +00:00
} catch (const std::exception & err) {
add log_callback to llama_context_params for custom logging. (#2234) * add log_callback to llama_context_params for custom logging. * Fix macro expansion on gcc * Add struct llama_state for global variables and move log_callback there * Turn log level into enum and some minor changes. * Remove model_for_logging parameter (not needed anymore) * Convert remaining fprintf(stderr, ...) calls to use new macros. * Fix enum and initialize g_state * Fix log calls after merge * Fix missing static * Add back all the new lines in the logging strings * Add comment for llama_log_callback and replace remaining printf calls --------- Co-authored-by: grahameth <-> Co-authored-by: Helmut <helmut.buhler@inf.h-brs.de>
2023-08-09 20:46:40 +00:00
LLAMA_LOG_ERROR("%s: failed to apply lora adapter: %s\n", __func__, err.what());
Add LoRA support (#820)
2023-04-17 15:28:55 +00:00
return 1;
}
}
llama : add support for control vectors (#5970) * control vector api and implementation * control-vectors : minor code style updates * disable control vector when data == nullptr use -1 for disabled range (also on init) in case we ever support controlling layer 0 (embeddings) --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-15 20:43:02 +00:00
static bool llama_control_vector_init(struct llama_control_vector & cvec, const llama_model & model) {
GGML_ASSERT(cvec.tensors.empty());
GGML_ASSERT(cvec.ctxs.empty());
GGML_ASSERT(cvec.bufs.empty());
// count layer buffer types
std::map<ggml_backend_buffer_type_t, int> buft_layer_count;
for (int64_t i = 0; i < model.hparams.n_layer; i++) {
buft_layer_count[model.buft_layer[i].buft]++;
}
// allocate contexts
std::map<ggml_backend_buffer_type_t, ggml_context *> ctx_map;
for (auto & it : buft_layer_count) {
int n_layers = it.second;
struct ggml_init_params params = {
/*.mem_size =*/ n_layers * ggml_tensor_overhead(),
/*.mem_buffer =*/ NULL,
/*.no_alloc =*/ true,
};
ggml_context * ctx = ggml_init(params);
if (!ctx) {
LLAMA_LOG_ERROR("%s: failed to allocate context for control vector\n", __func__);
return 1;
}
ctx_map[it.first] = ctx;
}
// make tensors
cvec.tensors.push_back(nullptr); // there's never a tensor for layer 0
for (size_t il = 1; il < model.hparams.n_layer; il++) {
struct ggml_context * ctx = ctx_map.at(model.buft_layer[il].buft);
ggml_tensor * tensor = ggml_new_tensor_1d(ctx, GGML_TYPE_F32, model.hparams.n_embd);
cvec.tensors.push_back(tensor);
}
// allocate tensors / buffers and zero
for (auto it : ctx_map) {
ggml_backend_buffer_type_t buft = it.first;
ggml_context * ctx = it.second;
ggml_backend_buffer_t buf = ggml_backend_alloc_ctx_tensors_from_buft(ctx, buft);
if (!buf) {
LLAMA_LOG_ERROR("%s: failed to allocate buffer for control vector\n", __func__);
return false;
}
ggml_backend_buffer_clear(buf, 0);
cvec.ctxs.push_back(ctx);
cvec.bufs.push_back(buf);
}
return true;
}
int32_t llama_control_vector_apply(struct llama_context * lctx, const float * data, size_t len, int32_t n_embd, int32_t il_start, int32_t il_end) {
const llama_model & model = lctx->model;
llama_control_vector & cvec = lctx->cvec;
if (data == nullptr) {
// disable the current control vector (but leave allocated for later)
cvec.layer_start = -1;
cvec.layer_end = -1;
return 0;
}
if (n_embd != (int) model.hparams.n_embd) {
LLAMA_LOG_ERROR("%s: control vector n_embd does not match model\n", __func__);
return 1;
}
if (cvec.tensors.empty()) {
if (!llama_control_vector_init(cvec, model)) {
return 1;
}
}
cvec.layer_start = il_start;
cvec.layer_end = il_end;
for (size_t il = 1; il < model.hparams.n_layer; il++) {
assert(cvec.tensors[il] != nullptr);
const size_t off = n_embd * (il - 1); // buffer doesn't have data for layer 0, since it's never present
if (off + n_embd <= len) {
ggml_backend_tensor_set(cvec.tensors[il], data + off, 0, n_embd * ggml_element_size(cvec.tensors[il]));
}
}
return 0;
}
llama : more consistent names of count variables (#5994) * llama : more consistent names of count variables ggml-ci * llama : n_parallel -> n_seq_max * common : fix param name * examples : fix param name
2024-03-11 15:49:47 +00:00
struct llama_kv_cache_view llama_kv_cache_view_init(const struct llama_context * ctx, int32_t n_seq_max) {
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
struct llama_kv_cache_view result = {
/*.n_cells = */ 0,
llama : more consistent names of count variables (#5994) * llama : more consistent names of count variables ggml-ci * llama : n_parallel -> n_seq_max * common : fix param name * examples : fix param name
2024-03-11 15:49:47 +00:00
/*.n_seq_max = */ n_seq_max,
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
/*.token_count = */ 0,
/*.used_cells = */ llama_get_kv_cache_used_cells(ctx),
/*.max_contiguous = */ 0,
/*.max_contiguous_idx = */ -1,
/*.cells = */ nullptr,
/*.cells_sequences = */ nullptr,
};
return result;
}
void llama_kv_cache_view_free(struct llama_kv_cache_view * view) {
if (view->cells != nullptr) {
free(view->cells);
view->cells = nullptr;
}
if (view->cells_sequences != nullptr) {
free(view->cells_sequences);
view->cells_sequences = nullptr;
}
}
void llama_kv_cache_view_update(const struct llama_context * ctx, struct llama_kv_cache_view * view) {
if (uint32_t(view->n_cells) < ctx->kv_self.size || view->cells == nullptr) {
view->n_cells = int32_t(ctx->kv_self.size);
void * p = realloc(view->cells, sizeof(struct llama_kv_cache_view_cell) * view->n_cells);
GGML_ASSERT(p != nullptr && "Failed to alloc kv_cache_view cells");
view->cells = (struct llama_kv_cache_view_cell *)p;
llama : more consistent names of count variables (#5994) * llama : more consistent names of count variables ggml-ci * llama : n_parallel -> n_seq_max * common : fix param name * examples : fix param name
2024-03-11 15:49:47 +00:00
p = realloc(view->cells_sequences, sizeof(llama_seq_id) * view->n_seq_max * view->n_cells);
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
GGML_ASSERT(p != nullptr && "Failed to alloc kv_cache_view cells sequences");
view->cells_sequences = (llama_seq_id *)p;
}
const std::vector<llama_kv_cell> & kv_cells = ctx->kv_self.cells;
llama_kv_cache_view_cell * c_curr = view->cells;
llama_seq_id * cs_curr = view->cells_sequences;
int32_t used_cells = 0;
int32_t token_count = 0;
int32_t curr_contig_idx = -1;
uint32_t max_contig = 0;
int32_t max_contig_idx = -1;
llama : more consistent names of count variables (#5994) * llama : more consistent names of count variables ggml-ci * llama : n_parallel -> n_seq_max * common : fix param name * examples : fix param name
2024-03-11 15:49:47 +00:00
for (int32_t i = 0; i < int32_t(ctx->kv_self.size); i++, c_curr++, cs_curr += view->n_seq_max) {
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
const size_t curr_size = kv_cells[i].seq_id.size();
token_count += curr_size;
c_curr->pos = kv_cells[i].pos + kv_cells[i].delta;
if (curr_size > 0) {
if (curr_contig_idx >= 0 && uint32_t(i - curr_contig_idx) > max_contig) {
max_contig = i - curr_contig_idx;
max_contig_idx = curr_contig_idx;
}
curr_contig_idx = -1;
} else if (curr_contig_idx < 0) {
curr_contig_idx = i;
}
int seq_idx = 0;
for (const llama_seq_id it : kv_cells[i].seq_id) {
llama : more consistent names of count variables (#5994) * llama : more consistent names of count variables ggml-ci * llama : n_parallel -> n_seq_max * common : fix param name * examples : fix param name
2024-03-11 15:49:47 +00:00
if (seq_idx >= view->n_seq_max) {
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
break;
}
cs_curr[seq_idx] = it;
seq_idx++;
}
if (seq_idx != 0) {
used_cells++;
}
llama : more consistent names of count variables (#5994) * llama : more consistent names of count variables ggml-ci * llama : n_parallel -> n_seq_max * common : fix param name * examples : fix param name
2024-03-11 15:49:47 +00:00
for (; seq_idx < view->n_seq_max; seq_idx++) {
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
cs_curr[seq_idx] = -1;
}
}
if (curr_contig_idx >= 0 && kv_cells.size() - curr_contig_idx > max_contig) {
max_contig_idx = curr_contig_idx;
max_contig = kv_cells.size() - curr_contig_idx;
}
view->max_contiguous = max_contig;
view->max_contiguous_idx = max_contig_idx;
view->token_count = token_count;
view->used_cells = used_cells;
if (uint32_t(used_cells) != ctx->kv_self.used) {
LLAMA_LOG_ERROR("%s: used cells mismatch. kv_cache says %d but we calculated %d\n",
__func__, ctx->kv_self.used, used_cells);
}
}
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
int32_t llama_get_kv_cache_token_count(const struct llama_context * ctx) {
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
int result = 0;
for (uint32_t i = 0; i < ctx->kv_self.size; i++) {
result += ctx->kv_self.cells[i].seq_id.size();
}
return result;
}
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
int32_t llama_get_kv_cache_used_cells(const struct llama_context * ctx) {
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
return ctx->kv_self.used;
Added api for getting/setting the kv_cache (#685) The api provides access methods for retrieving the current memory buffer for the kv_cache and its token number. It also contains a method for setting the kv_cache from a memory buffer. This makes it possible to load/save history - maybe support --cache-prompt paramater as well? Co-authored-by: Pavol Rusnak <pavol@rusnak.io>
2023-04-02 10:23:04 +00:00
}
Extend llama_kv_cache_seq_rm to allow matching any sequence (#3843) * Extend llama_kv_cache_seq_rm to allow matichng any sequence * Replace llama_kv_cache_tokens_rm with llama_kv_cache_clear Use llama_kv_cache_clear for cache clearing Change calls to llama_kv_cache_tokens_rm that want to delete by position to use llama_kv_cache_seq_rm functionality
2023-10-29 17:31:40 +00:00
void llama_kv_cache_clear(struct llama_context * ctx) {
llama_kv_cache_clear(ctx->kv_self);
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
}
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
bool llama_kv_cache_seq_rm(struct llama_context * ctx, llama_seq_id seq_id, llama_pos p0, llama_pos p1) {
return llama_kv_cache_seq_rm(ctx->kv_self, seq_id, p0, p1);
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
}
void llama_kv_cache_seq_cp(struct llama_context * ctx, llama_seq_id seq_id_src, llama_seq_id seq_id_dst, llama_pos p0, llama_pos p1) {
speculative : add tree-based sampling example (#3624) * sampling : one sequence per sampling context ggml-ci * speculative : add tree-based sampling support ggml-ci * speculative : reuse the n_parallel CLI param * speculative : refactor sampling * examples : fix build after sampling refactoring ggml-ci * batched : fix n_seq_id * sampling : fix malloc ggml-ci * swift : fix build ggml-ci * swift : try to fix build ggml-ci * prompts : add assistant.txt * common : add llama_batch_add() and llama_batch_clear() helpers * speculative : minor refactor ggml-ci * minor : comments + rename ggml-ci * speculative : fix off-by-one for n_drafted * speculative : fix the n_drafted fix + p constants
2023-10-18 13:21:57 +00:00
if (seq_id_src == seq_id_dst) {
return;
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
llama_kv_cache_seq_cp(ctx->kv_self, seq_id_src, seq_id_dst, p0, p1);
}
void llama_kv_cache_seq_keep(struct llama_context * ctx, llama_seq_id seq_id) {
llama_kv_cache_seq_keep(ctx->kv_self, seq_id);
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
void llama_kv_cache_seq_add(struct llama_context * ctx, llama_seq_id seq_id, llama_pos p0, llama_pos p1, llama_pos delta) {
examples : add passkey test (#3856) * examples : add passkey test * passkey : better prints * passkey : select pass key pos from CLI * passkey : simplify n_past logic * make : add passkey target * passkey : add "self-extend"-like context extension (#4810) * llama : "self-extend"-like context extension * passkey : add comment * passkey : add readme
2024-01-08 09:14:04 +00:00
if (delta == 0) {
return;
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
llama_kv_cache_seq_add(ctx->kv_self, seq_id, p0, p1, delta);
Allow setting the rng seed after initialization. (#1184) The llama_set_state_data function restores the rng state to what it was at the time llama_copy_state_data was called. But users may want to restore the state and proceed with a different seed.
2023-04-26 20:08:43 +00:00
}
examples : add passkey test (#3856) * examples : add passkey test * passkey : better prints * passkey : select pass key pos from CLI * passkey : simplify n_past logic * make : add passkey target * passkey : add "self-extend"-like context extension (#4810) * llama : "self-extend"-like context extension * passkey : add comment * passkey : add readme
2024-01-08 09:14:04 +00:00
void llama_kv_cache_seq_div(struct llama_context * ctx, llama_seq_id seq_id, llama_pos p0, llama_pos p1, int d) {
if (d == 1) {
return;
}
llama_kv_cache_seq_div(ctx->kv_self, seq_id, p0, p1, d);
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
llama_pos llama_kv_cache_seq_pos_max(struct llama_context * ctx, llama_seq_id seq_id) {
return llama_kv_cache_seq_pos_max(ctx->kv_self, seq_id);
}
void llama_kv_cache_defrag(struct llama_context * ctx) {
llama_kv_cache_defrag(ctx->kv_self);
}
void llama_kv_cache_update(struct llama_context * ctx) {
llama_kv_cache_update_internal(*ctx);
}
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
// deprecated
size_t llama_get_state_size(const struct llama_context * ctx) {
return llama_state_get_size(ctx);
}
// deprecated
size_t llama_copy_state_data(struct llama_context * ctx, uint8_t * dst) {
return llama_state_get_data(ctx, dst);
}
// deprecated
size_t llama_set_state_data(struct llama_context * ctx, const uint8_t * src) {
return llama_state_set_data(ctx, src);
}
// deprecated
bool llama_load_session_file(struct llama_context * ctx, const char * path_session, llama_token * tokens_out, size_t n_token_capacity, size_t * n_token_count_out) {
return llama_state_load_file(ctx, path_session, tokens_out, n_token_capacity, n_token_count_out);
}
// deprecated
bool llama_save_session_file(struct llama_context * ctx, const char * path_session, const llama_token * tokens, size_t n_token_count) {
return llama_state_save_file(ctx, path_session, tokens, n_token_count);
}
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
llama : only copy used KV cache in get / set state (#1272) * llama : only copy used KV cache in get / set state * switch to ggml for copying k, v * avoid designated initializers
2023-05-03 02:26:13 +00:00
// Returns the *maximum* size of the state
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
size_t llama_state_get_size(const struct llama_context * ctx) {
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
const auto & cparams = ctx->cparams;
const auto & hparams = ctx->model.hparams;
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
// we don't know size of rng until we actually serialize it. so reserve more than enough memory for its serialized state.
// for reference, std::mt19937(1337) serializes to 6701 bytes.
const size_t s_rng_size = sizeof(size_t);
const size_t s_rng = LLAMA_MAX_RNG_STATE;
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
const size_t s_n_outputs = sizeof(size_t);
// assume worst case for outputs although only currently set ones are serialized
const size_t s_output_pos = ctx->cparams.n_batch * sizeof(int32_t);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
const size_t s_logits_size = sizeof(size_t);
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
const size_t s_logits = ctx->logits_size ? cparams.n_batch * hparams.n_vocab * sizeof(float) : 0;
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
const size_t s_embedding_size = sizeof(size_t);
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
const size_t s_embedding = ctx->embd_size ? cparams.n_batch * hparams.n_embd * sizeof(float) : 0;
llama : fix llama_copy_state_data with fragmented KV cache (#5840) The row size of the saved states was based on kv_self.head while it should be based on llama_kv_cache_cell_max. Existing session files should still work. * llama : fix llama_kv_cache_cell_max inability to return 1 I've also changed its return type to uint32_t, because this function is always used to set the value of uint32_t variables, and because the index already has this type. * llama : fix state size calculation Some bytes in the state were unaccounted for in llama_get_state_size. Since the logits reserve so much space, it did not cause problems.
2024-03-03 08:41:55 +00:00
const size_t s_kv_buf_size = sizeof(size_t);
const size_t s_kv_head = sizeof(uint32_t);
const size_t s_kv_size = sizeof(uint32_t);
const size_t s_kv_used = sizeof(uint32_t);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
const size_t s_v_trans = sizeof(uint32_t);
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
const size_t s_kv = ctx->kv_self.total_size();
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
const size_t s_kv_cell = sizeof(llama_pos) + sizeof(size_t) + cparams.n_seq_max*sizeof(llama_seq_id);
llama : fix llama_copy_state_data with fragmented KV cache (#5840) The row size of the saved states was based on kv_self.head while it should be based on llama_kv_cache_cell_max. Existing session files should still work. * llama : fix llama_kv_cache_cell_max inability to return 1 I've also changed its return type to uint32_t, because this function is always used to set the value of uint32_t variables, and because the index already has this type. * llama : fix state size calculation Some bytes in the state were unaccounted for in llama_get_state_size. Since the logits reserve so much space, it did not cause problems.
2024-03-03 08:41:55 +00:00
const size_t s_kv_cells = ctx->kv_self.size * s_kv_cell;
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
const size_t s_total = (
+ s_rng_size
+ s_rng
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
+ s_n_outputs
+ s_output_pos
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
+ s_logits_size
+ s_logits
+ s_embedding_size
+ s_embedding
llama : fix llama_copy_state_data with fragmented KV cache (#5840) The row size of the saved states was based on kv_self.head while it should be based on llama_kv_cache_cell_max. Existing session files should still work. * llama : fix llama_kv_cache_cell_max inability to return 1 I've also changed its return type to uint32_t, because this function is always used to set the value of uint32_t variables, and because the index already has this type. * llama : fix state size calculation Some bytes in the state were unaccounted for in llama_get_state_size. Since the logits reserve so much space, it did not cause problems.
2024-03-03 08:41:55 +00:00
+ s_kv_buf_size
+ s_kv_head
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
+ s_kv_size
llama : fix llama_copy_state_data with fragmented KV cache (#5840) The row size of the saved states was based on kv_self.head while it should be based on llama_kv_cache_cell_max. Existing session files should still work. * llama : fix llama_kv_cache_cell_max inability to return 1 I've also changed its return type to uint32_t, because this function is always used to set the value of uint32_t variables, and because the index already has this type. * llama : fix state size calculation Some bytes in the state were unaccounted for in llama_get_state_size. Since the logits reserve so much space, it did not cause problems.
2024-03-03 08:41:55 +00:00
+ s_kv_used
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
+ s_v_trans
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
+ s_kv
llama : fix llama_copy_state_data with fragmented KV cache (#5840) The row size of the saved states was based on kv_self.head while it should be based on llama_kv_cache_cell_max. Existing session files should still work. * llama : fix llama_kv_cache_cell_max inability to return 1 I've also changed its return type to uint32_t, because this function is always used to set the value of uint32_t variables, and because the index already has this type. * llama : fix state size calculation Some bytes in the state were unaccounted for in llama_get_state_size. Since the logits reserve so much space, it did not cause problems.
2024-03-03 08:41:55 +00:00
+ s_kv_cells
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
// on session change it is very likely that the state size has changed - so we need to update this function
static_assert(LLAMA_SESSION_VERSION == 6, "So you just bumped the session version - good. But did you remember to update llama_state_get_size?");
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
return s_total;
Added api for getting/setting the kv_cache (#685) The api provides access methods for retrieving the current memory buffer for the kv_cache and its token number. It also contains a method for setting the kv_cache from a memory buffer. This makes it possible to load/save history - maybe support --cache-prompt paramater as well? Co-authored-by: Pavol Rusnak <pavol@rusnak.io>
2023-04-02 10:23:04 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
// llama_context_data
struct llama_data_context {
virtual void write(const void * src, size_t size) = 0;
virtual size_t get_size_written() = 0;
virtual ~llama_data_context() = default;
};
struct llama_data_buffer_context : llama_data_context {
uint8_t * ptr;
size_t size_written = 0;
llama_data_buffer_context(uint8_t * p) : ptr(p) {}
void write(const void * src, size_t size) override {
memcpy(ptr, src, size);
ptr += size;
size_written += size;
}
size_t get_size_written() override {
return size_written;
}
};
struct llama_data_file_context : llama_data_context {
llama_file * file;
size_t size_written = 0;
llama_data_file_context(llama_file * f) : file(f) {}
void write(const void * src, size_t size) override {
file->write_raw(src, size);
size_written += size;
}
size_t get_size_written() override {
return size_written;
}
};
Stream save llama context data to file instead of allocating entire buffer upfront (#2488) * added stream saving context data to file to avoid allocating unnecessary amounts of memory * generalised copying state data to file or buffer * added comments explaining how copy_state_data works * fixed trailing whitespaces * fixed save load state example * updated save load state to use public function in llama.cpp * - restored breakage of the llama_copy_state_data API - moved new logic for copying llama state data to internal function * fixed function declaration order * restored save load state example * fixed whitepace * removed unused llama-util.h include * Apply suggestions from code review Co-authored-by: slaren <slarengh@gmail.com> * Apply code review suggestions Co-authored-by: slaren <slarengh@gmail.com> --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-08-04 11:29:52 +00:00
/** copy state data into either a buffer or file depending on the passed in context
*
* file context:
* llama_file file("/path", "wb");
* llama_data_file_context data_ctx(&file);
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
* llama_state_get_data(ctx, &data_ctx);
Stream save llama context data to file instead of allocating entire buffer upfront (#2488) * added stream saving context data to file to avoid allocating unnecessary amounts of memory * generalised copying state data to file or buffer * added comments explaining how copy_state_data works * fixed trailing whitespaces * fixed save load state example * updated save load state to use public function in llama.cpp * - restored breakage of the llama_copy_state_data API - moved new logic for copying llama state data to internal function * fixed function declaration order * restored save load state example * fixed whitepace * removed unused llama-util.h include * Apply suggestions from code review Co-authored-by: slaren <slarengh@gmail.com> * Apply code review suggestions Co-authored-by: slaren <slarengh@gmail.com> --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-08-04 11:29:52 +00:00
*
* buffer context:
* std::vector<uint8_t> buf(max_size, 0);
* llama_data_buffer_context data_ctx(&buf.data());
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
* llama_state_get_data(ctx, &data_ctx);
Stream save llama context data to file instead of allocating entire buffer upfront (#2488) * added stream saving context data to file to avoid allocating unnecessary amounts of memory * generalised copying state data to file or buffer * added comments explaining how copy_state_data works * fixed trailing whitespaces * fixed save load state example * updated save load state to use public function in llama.cpp * - restored breakage of the llama_copy_state_data API - moved new logic for copying llama state data to internal function * fixed function declaration order * restored save load state example * fixed whitepace * removed unused llama-util.h include * Apply suggestions from code review Co-authored-by: slaren <slarengh@gmail.com> * Apply code review suggestions Co-authored-by: slaren <slarengh@gmail.com> --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-08-04 11:29:52 +00:00
*
*/
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
static void llama_state_get_data_internal(struct llama_context * ctx, llama_data_context * data_ctx) {
llama : synchronize before get/set session data (#6911)
2024-04-25 15:59:03 +00:00
llama_synchronize(ctx);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
// copy rng
{
llama : minimize size used for state save/load (#4820) * examples : save-load-state: save only required state * llama : only reserve n_vocab * n_batch at most for logits llama_decode asserts that only n_batch tokens are passed each call, and n_ctx is expected to be bigger than n_batch. * llama : always reserve n_vocab * n_batch for logits llama_context de-serialization breaks if the contexts have differing capacity for logits and llama_decode will at maximum resize to n_vocab * n_batch. * llama : only save and restore used logits for batch sizes of 512 this reduces save state in the best case by around 62 MB, which can be a lot if planning to save on each message to allow regenerating messages. * llama : use ostringstream and istringstream for save and load * llama : serialize rng into minimum amount of space required * llama : break session version due to serialization changes
2024-01-13 16:29:43 +00:00
std::ostringstream rng_ss;
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
rng_ss << ctx->rng;
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
const std::string & rng_str = rng_ss.str();
llama : minimize size used for state save/load (#4820) * examples : save-load-state: save only required state * llama : only reserve n_vocab * n_batch at most for logits llama_decode asserts that only n_batch tokens are passed each call, and n_ctx is expected to be bigger than n_batch. * llama : always reserve n_vocab * n_batch for logits llama_context de-serialization breaks if the contexts have differing capacity for logits and llama_decode will at maximum resize to n_vocab * n_batch. * llama : only save and restore used logits for batch sizes of 512 this reduces save state in the best case by around 62 MB, which can be a lot if planning to save on each message to allow regenerating messages. * llama : use ostringstream and istringstream for save and load * llama : serialize rng into minimum amount of space required * llama : break session version due to serialization changes
2024-01-13 16:29:43 +00:00
const size_t rng_size = rng_str.size();
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
llama : minimize size used for state save/load (#4820) * examples : save-load-state: save only required state * llama : only reserve n_vocab * n_batch at most for logits llama_decode asserts that only n_batch tokens are passed each call, and n_ctx is expected to be bigger than n_batch. * llama : always reserve n_vocab * n_batch for logits llama_context de-serialization breaks if the contexts have differing capacity for logits and llama_decode will at maximum resize to n_vocab * n_batch. * llama : only save and restore used logits for batch sizes of 512 this reduces save state in the best case by around 62 MB, which can be a lot if planning to save on each message to allow regenerating messages. * llama : use ostringstream and istringstream for save and load * llama : serialize rng into minimum amount of space required * llama : break session version due to serialization changes
2024-01-13 16:29:43 +00:00
GGML_ASSERT(rng_size <= LLAMA_MAX_RNG_STATE);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
llama : minimize size used for state save/load (#4820) * examples : save-load-state: save only required state * llama : only reserve n_vocab * n_batch at most for logits llama_decode asserts that only n_batch tokens are passed each call, and n_ctx is expected to be bigger than n_batch. * llama : always reserve n_vocab * n_batch for logits llama_context de-serialization breaks if the contexts have differing capacity for logits and llama_decode will at maximum resize to n_vocab * n_batch. * llama : only save and restore used logits for batch sizes of 512 this reduces save state in the best case by around 62 MB, which can be a lot if planning to save on each message to allow regenerating messages. * llama : use ostringstream and istringstream for save and load * llama : serialize rng into minimum amount of space required * llama : break session version due to serialization changes
2024-01-13 16:29:43 +00:00
data_ctx->write(&rng_size, sizeof(rng_size));
data_ctx->write(rng_str.data(), rng_size);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
// copy outputs
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
{
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
// Can't use ctx->n_outputs because it's not for the
// entire last batch when n_ubatch is smaller than n_batch
size_t n_outputs = 0;
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
// copy output ids
{
std::vector<int32_t> output_pos;
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
const size_t n_batch = ctx->cparams.n_batch;
const auto & output_ids = ctx->output_ids;
output_pos.resize(ctx->output_size);
// build a more compact representation of the output ids
for (size_t i = 0; i < n_batch; ++i) {
// map an output id to a position in the batch
int32_t pos = output_ids[i];
if (pos >= 0) {
if ((size_t) pos >= n_outputs) {
n_outputs = pos + 1;
}
GGML_ASSERT((size_t) pos < ctx->output_size);
output_pos[pos] = i;
}
}
data_ctx->write(&n_outputs, sizeof(n_outputs));
if (n_outputs) {
data_ctx->write(output_pos.data(), n_outputs * sizeof(int32_t));
}
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
// copy logits
{
const size_t logits_size = std::min(ctx->logits_size, n_outputs * ctx->model.hparams.n_vocab);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
data_ctx->write(&logits_size, sizeof(logits_size));
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (logits_size) {
data_ctx->write(ctx->logits, logits_size * sizeof(float));
}
}
// copy embeddings
{
const size_t embeddings_size = std::min(ctx->embd_size, n_outputs * ctx->model.hparams.n_embd);
data_ctx->write(&embeddings_size, sizeof(embeddings_size));
if (embeddings_size) {
data_ctx->write(ctx->embd, embeddings_size * sizeof(float));
}
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
}
}
// copy kv cache
{
llama : make model stateless and context stateful (llama_state) (#1797) * llama : make model stateless and context stateful * llama : minor cleanup * llama : update internal API declaration * Apply suggestions from code review fix style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Missing model memory release * Fix style * Add deprecated warning for public API function llama_init_from_file * Update public API use cases: move away from deprecated llama_init_from_file * Deprecate public API function llama_apply_lora_from_file --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-24 08:47:58 +00:00
const auto & kv_self = ctx->kv_self;
llama : only copy used KV cache in get / set state (#1272) * llama : only copy used KV cache in get / set state * switch to ggml for copying k, v * avoid designated initializers
2023-05-03 02:26:13 +00:00
const auto & hparams = ctx->model.hparams;
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
const uint32_t n_layer = hparams.n_layer;
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa() + hparams.n_embd_k_s();
const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa() + hparams.n_embd_v_s();
llama : only copy used KV cache in get / set state (#1272) * llama : only copy used KV cache in get / set state * switch to ggml for copying k, v * avoid designated initializers
2023-05-03 02:26:13 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
// NOTE: kv_size and kv_buf_size are mostly used for sanity checks
llama : fix llama_copy_state_data with fragmented KV cache (#5840) The row size of the saved states was based on kv_self.head while it should be based on llama_kv_cache_cell_max. Existing session files should still work. * llama : fix llama_kv_cache_cell_max inability to return 1 I've also changed its return type to uint32_t, because this function is always used to set the value of uint32_t variables, and because the index already has this type. * llama : fix state size calculation Some bytes in the state were unaccounted for in llama_get_state_size. Since the logits reserve so much space, it did not cause problems.
2024-03-03 08:41:55 +00:00
const uint32_t kv_head = llama_kv_cache_cell_max(kv_self);
llama : fix session saving/loading (#3400) * llama : fix session saving/loading * llama : temp fix for clearing "future" tokens from the KV cache * llama : fix handling of "future" tokens when loading sessions * llama : fix comments for llama_kv_cache API
2023-10-03 18:04:01 +00:00
const uint32_t kv_size = kv_self.size;
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
const size_t kv_buf_size = kv_self.total_size() / (kv_size ? kv_size : 1) * kv_head;
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
const uint32_t kv_used = kv_self.used;
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
const uint32_t v_trans = kv_self.v_trans ? 1 : 0;
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
llama : fix session saving/loading (#3400) * llama : fix session saving/loading * llama : temp fix for clearing "future" tokens from the KV cache * llama : fix handling of "future" tokens when loading sessions * llama : fix comments for llama_kv_cache API
2023-10-03 18:04:01 +00:00
data_ctx->write(&kv_buf_size, sizeof(kv_buf_size));
data_ctx->write(&kv_head, sizeof(kv_head));
data_ctx->write(&kv_size, sizeof(kv_size));
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
data_ctx->write(&kv_used, sizeof(kv_used));
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
data_ctx->write(&v_trans, sizeof(v_trans));
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
llama : fix session saving/loading (#3400) * llama : fix session saving/loading * llama : temp fix for clearing "future" tokens from the KV cache * llama : fix handling of "future" tokens when loading sessions * llama : fix comments for llama_kv_cache API
2023-10-03 18:04:01 +00:00
if (kv_buf_size) {
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
const size_t pre_kv_buf_size = data_ctx->get_size_written();
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
std::vector<uint8_t> tmp_buf;
for (int il = 0; il < (int) n_layer; ++il) {
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
const size_t k_size = ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa*kv_head);
llama : fix session save/load with quantized KV (#5649)
2024-02-21 21:52:39 +00:00
tmp_buf.resize(k_size);
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ggml_backend_tensor_get(kv_self.k_l[il], tmp_buf.data(), 0, tmp_buf.size());
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
data_ctx->write(tmp_buf.data(), tmp_buf.size());
Stream save llama context data to file instead of allocating entire buffer upfront (#2488) * added stream saving context data to file to avoid allocating unnecessary amounts of memory * generalised copying state data to file or buffer * added comments explaining how copy_state_data works * fixed trailing whitespaces * fixed save load state example * updated save load state to use public function in llama.cpp * - restored breakage of the llama_copy_state_data API - moved new logic for copying llama state data to internal function * fixed function declaration order * restored save load state example * fixed whitepace * removed unused llama-util.h include * Apply suggestions from code review Co-authored-by: slaren <slarengh@gmail.com> * Apply code review suggestions Co-authored-by: slaren <slarengh@gmail.com> --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-08-04 11:29:52 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
if (kv_self.recurrent || !kv_self.v_trans) {
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
// v is contiguous for recurrent models
// TODO: use other tensors for state models than k and v
const size_t v_size = ggml_row_size(kv_self.v_l[il]->type, n_embd_v_gqa*kv_head);
tmp_buf.resize(v_size);
ggml_backend_tensor_get(kv_self.v_l[il], tmp_buf.data(), 0, tmp_buf.size());
data_ctx->write(tmp_buf.data(), tmp_buf.size());
continue;
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// v is not contiguous, copy row by row
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
const size_t v_row_size = ggml_row_size(kv_self.v_l[il]->type, kv_head);
llama : fix llama_copy_state_data with fragmented KV cache (#5840) The row size of the saved states was based on kv_self.head while it should be based on llama_kv_cache_cell_max. Existing session files should still work. * llama : fix llama_kv_cache_cell_max inability to return 1 I've also changed its return type to uint32_t, because this function is always used to set the value of uint32_t variables, and because the index already has this type. * llama : fix state size calculation Some bytes in the state were unaccounted for in llama_get_state_size. Since the logits reserve so much space, it did not cause problems.
2024-03-03 08:41:55 +00:00
const size_t v_row_stride = ggml_row_size(kv_self.v_l[il]->type, kv_size);
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
llama : fix session save/load with quantized KV (#5649)
2024-02-21 21:52:39 +00:00
tmp_buf.resize(v_row_size);
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (int ir = 0; ir < (int) n_embd_v_gqa; ++ir) {
llama : fix session save/load with quantized KV (#5649)
2024-02-21 21:52:39 +00:00
ggml_backend_tensor_get(kv_self.v_l[il], tmp_buf.data(), ir*v_row_stride, tmp_buf.size());
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
data_ctx->write(tmp_buf.data(), tmp_buf.size());
}
llama : per-layer KV cache + quantum K cache (#4309) * per-layer KV * remove unnecessary copies * less code duplication, offload k and v separately * llama : offload KV cache per-layer * llama : offload K shift tensors * llama : offload for rest of the model arches * llama : enable offload debug temporarily * llama : keep the KV related layers on the device * llama : remove mirrors, perform Device -> Host when partial offload * common : add command-line arg to disable KV cache offloading * llama : update session save/load * llama : support quantum K cache (#4312) * llama : support quantum K cache (wip) * metal : add F32 -> Q8_0 copy kernel * cuda : add F32 -> Q8_0 copy kernel ggml-ci * cuda : use mmv kernel for quantum cache ops * llama : pass KV cache type through API * llama : fix build ggml-ci * metal : add F32 -> Q4_0 copy kernel * metal : add F32 -> Q4_1 copy kernel * cuda : wip * cuda : add F32 -> Q4_0 and F32 -> Q4_1 copy kernels * llama-bench : support type_k/type_v * metal : use mm kernel only for quantum KV cache * cuda : add comment * llama : remove memory_f16 and kv_f16 flags --------- Co-authored-by: slaren <slarengh@gmail.com> * readme : add API change notice --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-12-07 11:03:17 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
GGML_ASSERT(kv_buf_size == data_ctx->get_size_written() - pre_kv_buf_size);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
}
llama : fix session saving/loading (#3400) * llama : fix session saving/loading * llama : temp fix for clearing "future" tokens from the KV cache * llama : fix handling of "future" tokens when loading sessions * llama : fix comments for llama_kv_cache API
2023-10-03 18:04:01 +00:00
llama : fix llama_copy_state_data with fragmented KV cache (#5840) The row size of the saved states was based on kv_self.head while it should be based on llama_kv_cache_cell_max. Existing session files should still work. * llama : fix llama_kv_cache_cell_max inability to return 1 I've also changed its return type to uint32_t, because this function is always used to set the value of uint32_t variables, and because the index already has this type. * llama : fix state size calculation Some bytes in the state were unaccounted for in llama_get_state_size. Since the logits reserve so much space, it did not cause problems.
2024-03-03 08:41:55 +00:00
for (uint32_t i = 0; i < kv_head; ++i) {
llama : fix session saving/loading (#3400) * llama : fix session saving/loading * llama : temp fix for clearing "future" tokens from the KV cache * llama : fix handling of "future" tokens when loading sessions * llama : fix comments for llama_kv_cache API
2023-10-03 18:04:01 +00:00
const auto & cell = kv_self.cells[i];
const llama_pos pos = cell.pos;
const size_t seq_id_size = cell.seq_id.size();
data_ctx->write(&pos, sizeof(pos));
data_ctx->write(&seq_id_size, sizeof(seq_id_size));
for (auto seq_id : cell.seq_id) {
data_ctx->write(&seq_id, sizeof(seq_id));
}
}
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
}
Stream save llama context data to file instead of allocating entire buffer upfront (#2488) * added stream saving context data to file to avoid allocating unnecessary amounts of memory * generalised copying state data to file or buffer * added comments explaining how copy_state_data works * fixed trailing whitespaces * fixed save load state example * updated save load state to use public function in llama.cpp * - restored breakage of the llama_copy_state_data API - moved new logic for copying llama state data to internal function * fixed function declaration order * restored save load state example * fixed whitepace * removed unused llama-util.h include * Apply suggestions from code review Co-authored-by: slaren <slarengh@gmail.com> * Apply code review suggestions Co-authored-by: slaren <slarengh@gmail.com> --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-08-04 11:29:52 +00:00
}
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
size_t llama_state_get_data(struct llama_context * ctx, uint8_t * dst) {
Stream save llama context data to file instead of allocating entire buffer upfront (#2488) * added stream saving context data to file to avoid allocating unnecessary amounts of memory * generalised copying state data to file or buffer * added comments explaining how copy_state_data works * fixed trailing whitespaces * fixed save load state example * updated save load state to use public function in llama.cpp * - restored breakage of the llama_copy_state_data API - moved new logic for copying llama state data to internal function * fixed function declaration order * restored save load state example * fixed whitepace * removed unused llama-util.h include * Apply suggestions from code review Co-authored-by: slaren <slarengh@gmail.com> * Apply code review suggestions Co-authored-by: slaren <slarengh@gmail.com> --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-08-04 11:29:52 +00:00
llama_data_buffer_context data_ctx(dst);
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
llama_state_get_data_internal(ctx, &data_ctx);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
Stream save llama context data to file instead of allocating entire buffer upfront (#2488) * added stream saving context data to file to avoid allocating unnecessary amounts of memory * generalised copying state data to file or buffer * added comments explaining how copy_state_data works * fixed trailing whitespaces * fixed save load state example * updated save load state to use public function in llama.cpp * - restored breakage of the llama_copy_state_data API - moved new logic for copying llama state data to internal function * fixed function declaration order * restored save load state example * fixed whitepace * removed unused llama-util.h include * Apply suggestions from code review Co-authored-by: slaren <slarengh@gmail.com> * Apply code review suggestions Co-authored-by: slaren <slarengh@gmail.com> --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-08-04 11:29:52 +00:00
return data_ctx.get_size_written();
Added api for getting/setting the kv_cache (#685) The api provides access methods for retrieving the current memory buffer for the kv_cache and its token number. It also contains a method for setting the kv_cache from a memory buffer. This makes it possible to load/save history - maybe support --cache-prompt paramater as well? Co-authored-by: Pavol Rusnak <pavol@rusnak.io>
2023-04-02 10:23:04 +00:00
}
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
// Sets the state reading from the specified source address
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
size_t llama_state_set_data(struct llama_context * ctx, const uint8_t * src) {
llama : synchronize before get/set session data (#6911)
2024-04-25 15:59:03 +00:00
llama_synchronize(ctx);
llama : constified `llama_set_state_data`'s `src` (#5774)
2024-02-29 08:17:23 +00:00
const uint8_t * inp = src;
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
// set rng
{
size_t rng_size;
llama : minimize size used for state save/load (#4820) * examples : save-load-state: save only required state * llama : only reserve n_vocab * n_batch at most for logits llama_decode asserts that only n_batch tokens are passed each call, and n_ctx is expected to be bigger than n_batch. * llama : always reserve n_vocab * n_batch for logits llama_context de-serialization breaks if the contexts have differing capacity for logits and llama_decode will at maximum resize to n_vocab * n_batch. * llama : only save and restore used logits for batch sizes of 512 this reduces save state in the best case by around 62 MB, which can be a lot if planning to save on each message to allow regenerating messages. * llama : use ostringstream and istringstream for save and load * llama : serialize rng into minimum amount of space required * llama : break session version due to serialization changes
2024-01-13 16:29:43 +00:00
memcpy(&rng_size, inp, sizeof(rng_size)); inp += sizeof(rng_size);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
llama : minimize size used for state save/load (#4820) * examples : save-load-state: save only required state * llama : only reserve n_vocab * n_batch at most for logits llama_decode asserts that only n_batch tokens are passed each call, and n_ctx is expected to be bigger than n_batch. * llama : always reserve n_vocab * n_batch for logits llama_context de-serialization breaks if the contexts have differing capacity for logits and llama_decode will at maximum resize to n_vocab * n_batch. * llama : only save and restore used logits for batch sizes of 512 this reduces save state in the best case by around 62 MB, which can be a lot if planning to save on each message to allow regenerating messages. * llama : use ostringstream and istringstream for save and load * llama : serialize rng into minimum amount of space required * llama : break session version due to serialization changes
2024-01-13 16:29:43 +00:00
GGML_ASSERT(rng_size <= LLAMA_MAX_RNG_STATE);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
llama : constified `llama_set_state_data`'s `src` (#5774)
2024-02-29 08:17:23 +00:00
std::string rng_str((const char *)inp, rng_size); inp += rng_size;
llama : minimize size used for state save/load (#4820) * examples : save-load-state: save only required state * llama : only reserve n_vocab * n_batch at most for logits llama_decode asserts that only n_batch tokens are passed each call, and n_ctx is expected to be bigger than n_batch. * llama : always reserve n_vocab * n_batch for logits llama_context de-serialization breaks if the contexts have differing capacity for logits and llama_decode will at maximum resize to n_vocab * n_batch. * llama : only save and restore used logits for batch sizes of 512 this reduces save state in the best case by around 62 MB, which can be a lot if planning to save on each message to allow regenerating messages. * llama : use ostringstream and istringstream for save and load * llama : serialize rng into minimum amount of space required * llama : break session version due to serialization changes
2024-01-13 16:29:43 +00:00
std::istringstream rng_ss(rng_str);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
rng_ss >> ctx->rng;
fix some warnings from gcc and clang-tidy (#3038) Co-authored-by: xaedes <xaedes@gmail.com>
2023-09-07 17:22:29 +00:00
GGML_ASSERT(!rng_ss.fail());
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
// set output ids
{
size_t n_outputs;
std::vector<int32_t> output_pos;
memcpy(&n_outputs, inp, sizeof(n_outputs)); inp += sizeof(n_outputs);
GGML_ASSERT(n_outputs <= llama_output_reserve(*ctx, n_outputs));
if (n_outputs) {
output_pos.resize(n_outputs);
memcpy(output_pos.data(), inp, n_outputs * sizeof(int32_t));
inp += n_outputs * sizeof(int32_t);
for (int32_t i = 0; i < (int32_t) output_pos.size(); ++i) {
int32_t id = output_pos[i];
GGML_ASSERT((uint32_t) id < ctx->cparams.n_batch);
ctx->output_ids[id] = i;
}
llama : fix restoring the number of outputs from state files (#6687)
2024-04-15 12:56:55 +00:00
ctx->n_outputs = n_outputs;
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
}
}
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
// set logits
{
size_t logits_size;
llama : free ggml context in set / copy state data (close #1425)
2023-05-13 06:08:52 +00:00
memcpy(&logits_size, inp, sizeof(logits_size)); inp += sizeof(logits_size);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
GGML_ASSERT(ctx->logits_size >= logits_size);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
if (logits_size) {
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
memcpy(ctx->logits, inp, logits_size * sizeof(float));
llama : minimize size used for state save/load (#4820) * examples : save-load-state: save only required state * llama : only reserve n_vocab * n_batch at most for logits llama_decode asserts that only n_batch tokens are passed each call, and n_ctx is expected to be bigger than n_batch. * llama : always reserve n_vocab * n_batch for logits llama_context de-serialization breaks if the contexts have differing capacity for logits and llama_decode will at maximum resize to n_vocab * n_batch. * llama : only save and restore used logits for batch sizes of 512 this reduces save state in the best case by around 62 MB, which can be a lot if planning to save on each message to allow regenerating messages. * llama : use ostringstream and istringstream for save and load * llama : serialize rng into minimum amount of space required * llama : break session version due to serialization changes
2024-01-13 16:29:43 +00:00
inp += logits_size * sizeof(float);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
}
}
// set embeddings
{
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
size_t embeddings_size;
memcpy(&embeddings_size, inp, sizeof(embeddings_size)); inp += sizeof(embeddings_size);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
GGML_ASSERT(ctx->embd_size >= embeddings_size);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
if (embeddings_size) {
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
memcpy(ctx->embd, inp, embeddings_size * sizeof(float));
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
inp += embeddings_size * sizeof(float);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
}
}
// set kv cache
{
llama : make model stateless and context stateful (llama_state) (#1797) * llama : make model stateless and context stateful * llama : minor cleanup * llama : update internal API declaration * Apply suggestions from code review fix style Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Missing model memory release * Fix style * Add deprecated warning for public API function llama_init_from_file * Update public API use cases: move away from deprecated llama_init_from_file * Deprecate public API function llama_apply_lora_from_file --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-06-24 08:47:58 +00:00
const auto & kv_self = ctx->kv_self;
llama : only copy used KV cache in get / set state (#1272) * llama : only copy used KV cache in get / set state * switch to ggml for copying k, v * avoid designated initializers
2023-05-03 02:26:13 +00:00
const auto & hparams = ctx->model.hparams;
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
const uint32_t n_layer = hparams.n_layer;
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa() + hparams.n_embd_k_s();
const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa() + hparams.n_embd_v_s();
llama : only copy used KV cache in get / set state (#1272) * llama : only copy used KV cache in get / set state * switch to ggml for copying k, v * avoid designated initializers
2023-05-03 02:26:13 +00:00
llama : fix session saving/loading (#3400) * llama : fix session saving/loading * llama : temp fix for clearing "future" tokens from the KV cache * llama : fix handling of "future" tokens when loading sessions * llama : fix comments for llama_kv_cache API
2023-10-03 18:04:01 +00:00
size_t kv_buf_size;
uint32_t kv_head;
uint32_t kv_size;
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
uint32_t kv_used;
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
uint32_t v_trans;
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
llama : fix session saving/loading (#3400) * llama : fix session saving/loading * llama : temp fix for clearing "future" tokens from the KV cache * llama : fix handling of "future" tokens when loading sessions * llama : fix comments for llama_kv_cache API
2023-10-03 18:04:01 +00:00
memcpy(&kv_buf_size, inp, sizeof(kv_buf_size)); inp += sizeof(kv_buf_size);
memcpy(&kv_head, inp, sizeof(kv_head)); inp += sizeof(kv_head);
memcpy(&kv_size, inp, sizeof(kv_size)); inp += sizeof(kv_size);
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
memcpy(&kv_used, inp, sizeof(kv_used)); inp += sizeof(kv_used);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
memcpy(&v_trans, inp, sizeof(v_trans)); inp += sizeof(v_trans);
GGML_ASSERT(kv_self.v_trans == (bool) v_trans); // incompatible V transposition
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
if (kv_self.size != kv_size) {
// the KV cache needs to be big enough to load all the KV cells from the saved state
GGML_ASSERT(kv_self.size >= kv_head);
LLAMA_LOG_INFO("%s: state contains %d KV cells, was saved with kv_size=%d, but is loaded with kv_size=%d (fine, but different)\n",
__func__, kv_head, kv_size, kv_self.size);
}
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
llama_kv_cache_clear(ctx);
llama : fix session saving/loading (#3400) * llama : fix session saving/loading * llama : temp fix for clearing "future" tokens from the KV cache * llama : fix handling of "future" tokens when loading sessions * llama : fix comments for llama_kv_cache API
2023-10-03 18:04:01 +00:00
if (kv_buf_size) {
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
const size_t pre_kv_buf_size = inp - src;
GGML_ASSERT(kv_self.total_size() >= kv_buf_size);
llama : only copy used KV cache in get / set state (#1272) * llama : only copy used KV cache in get / set state * switch to ggml for copying k, v * avoid designated initializers
2023-05-03 02:26:13 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (int il = 0; il < (int) n_layer; ++il) {
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
const size_t k_size = ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa*kv_head);
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
ggml_backend_tensor_set(kv_self.k_l[il], inp, 0, k_size);
inp += k_size;
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
if (kv_self.recurrent || !kv_self.v_trans) {
llama : support Mamba Selective State Space Models (#5328) * mamba : begin working on support for Mamba SSM * mamba : begin figuring out how to (ab)use the kv cache for Mamba * mamba : recurrent inference almost works, but incoherent * mamba : recurrent inference WORKS!!! * convert : optionally use d_conv and d_state from config.json for Mamba * mamba : refactor recurrent conv, resulting in 20% perf increase It's still slower than I'd like, but I did not really optimize `ggml_exp` yet. I also refactored `ggml_exp` to work with tensors with more than 2 dimensions. * ggml : parallelize ggml_exp This results in 8% faster token generation for Mamba-130M. * mamba : simplify the conv step with a self-overlapping view Turns out the conv_state can be made smaller by one column. Note that this breaks existing GGUFs of Mamba, because the key_value_length field is tied to the conv_state size. Convolution with a self-overlapping view is cool! And it's much simpler than what I initially thought would be necessary to make the convolution step work with more than 1 token at a time. Next step is to make the SSM step work on batches of tokens too, and thus I need to figure out a way to make a parallel selective scan which will keep the ssm_state small and won't make it bigger by a factor of (n_layer * batch_size). * llama : fix Mamba KV self size wrongly displaying as f16 instead of f32 Relatedly, I also tried to see if other types than f32 worked for the states, but they don't, because of the operators used. It's probably better anyway to keep lots of precision there, since the states are small anyway. * mamba : fix self-overlapping view depth stride * mamba : handle batches of more than 1 token This means running Mamba no longer crashes when using the default settings! And probably also slightly faster prompt processing. Both batched and non-batched processing yield the same output. Previously, the state was not cleared when starting a sequence. Next step is to make the KV cache API work as expected for Mamba models. * ggml: add ggml_ssm_scan to help with parallel selective scan If the selective scan was implemented without a custom operator, there would be waaay too many nodes in the graph. For example, for Mamba-130M, with a batch size of 512 (the default), a naive selective scan could add at least 24*512=12288 nodes, which is more than LLAMA_MAX_NODES (8192), and that's only for the smallest Mamba model. So it's much cleaner with a custom operator. Not sure about the name, though. * ggml : in ggml_ssm_scan, merge multiple rows in the same vec operation This will help with performance on CPU if ggml_vec_mul_f32 and ggml_vec_add_f32 are ever optimized with SIMD. * mamba : very basic quantization support Mostly works, but there is currently no difference between the variants of a k-quant (e.g. Q4_K_S and Q4_K_M are the same). Most of the SSM-specific weights can be kept in f32 without affecting the size that much, since they are relatively small. (the linear projection weights are responsible for most of Mamba's size) Too much quantization seems to make the state degrade quite fast, and the model begins to output gibberish. It seems to affect bigger models to a lesser extent than small models, but I'm not sure by how much. Experimentation will be needed to figure out which weights are more important for the _M (and _L?) variants of k-quants for Mamba. * convert : fix wrong name for layer norm weight of offical Mamba models I was using Q-bert/Mamba-* models before, which have a slighlty different naming scheme for the weights. (they start with "model.layers" instead of "backbone.layers") * mamba : fuse more steps of the SSM scan in the ggml_ssm_scan operator This increases performance on CPU by around 30% for prompt processing, and by around 20% for text generation. However, it also makes the ggml_exp and ggml_soft_plus operators unused. Whether or not they should be kept will be decided later. * convert : for Mamba, also consider the "MambaLMHeadModel" arch name It's the name of the class of the official implementation, though they don't use it (yet) in the "architectures" field of config.json * mamba : fix vocab size problems with official models The perplexity was waaaay to high for models with a non-round vocab size. Not sure why, but it needed to be fixed in the metadata. Note that this breaks existing GGUF-converted Mamba models, but **only if** the vocab size was not already rounded. * ggml : remove ggml_exp and ggml_soft_plus They did not exist anyway outside of this branch, and since ggml_ssm_scan fused operations together, they are unused. It's always possible to bring them back if needed. * mamba : remove some useless comments No code change. * convert : fix flake8 linter errors * mamba : apply suggestions from code review * mamba : remove unecessary branch for row-wise ssm_state and C multiplication It was previously done to avoid permuting when only one token is processed at a time (like when generating text), but permuting is cheap, and dynamically changing the compute graph is not future-proof. * ggml : in ggml_ssm_scan, use more appropriate asserts * ggml : rename the destination pointer in ggml_compute_forward_ssm_scan_f32 * mamba : multiple sequences, but one at a time This is a step towards making this Mamba implementation usable with the server example (the way the system prompt is kept when clearing the client slots will need to be changed before this can work, though). The KV cache size for this kind of model is tied to the maximum number of sequences kept at any single time. For now, this number is obtained from n_parallel (plus one, to have an extra sequence to dedicate to the system prompt), but there might be a better way to do this which won't also make the main example use 2 cells even if only 1 is really used. (for this specific case, --parallel 0 helps) Simultaneous sequence processing will probably require changes to ggml_ssm_scan, and possibly a new operator for the conv step. * mamba : support llama_kv_cache_seq_cp This (mis)uses the logic around K shifts, because tokens in a state can't be shifted anyway, and because inp_K_shift has the right shape and type. Using ggml_get_rows is a nice way to do copies, but copy chains can't work. Fortunately, copy chains don't really seem to be used in the examples. Each KV cell is dedicated to the sequence ID corresponding to its own index. * mamba : use a state mask It's cleaner than the previous heuristic of checking for the pos of the first token in the batch. inp_KQ_mask could not be re-used for this, because it has the wrong shape and because it seems more suited to the next step of simultaneous sequence processing (helping with the problem of remembering which token belongs to which sequence(s)/state(s)). * llama : replace the usage of n_ctx with kv_self.size in many places * mamba : use n_tokens directly instead of n_tok * mamba : in comments, properly refer to KV cells instead of slots * mamba : reduce memory usage of ggml_ssm_scan From 290.37 MiB to 140.68 MiB of CPU compute buffer size with Mamba 3B with a batch size of 512. The result tensor of ggml_ssm_scan was previously a big part of the CPU compute buffer size. To make it smaller, it does not contain the intermediate ssm states anymore. Both y and the last ssm state are combined in the result tensor, because it seems only a single tensor can be returned by an operator with the way the graph is built. * mamba : simultaneous sequence processing A batch can now contain tokens from multiple sequences. This is necessary for at least the parallel example, the server example, and the HellaSwag test in the perplexity example. However, for this to be useful, uses of llama_kv_cache_seq_rm/cp will need to be changed to work on whole sequences. * ggml : add ggml_ssm_conv as a new operator for the conv step of Mamba This operator makes it possible to use and update the correct states for each token of the batch in the same way as ggml_ssm_scan. Other solutions which use existing operators would need loops which would add too many nodes to the graph (at least the ones I thought of). Using this operator further reduces the size of the CPU compute buffer from 140.68 MiB to 103.20 MiB with Mamba 3B with a batch size of 512. And (at least on CPU), it's a bit faster than before. Note that "ggml_ssm_conv" is probably not the most appropriate name, and it could be changed if a better one is found. * llama : add inp_s_seq as a new input tensor The most convenient implementation to select the correct state (for Mamba) for each token is to directly get the correct index from a tensor. This is why inp_s_seq is storing int32_t and not floats. The other, less convenient way to select the correct state would be to have inp_KQ_mask contain 1.0f for each state used by a token and 0.0f otherwise. This complicates quickly fetching the first used state of a token, and is also less efficient because a whole row of the mask would always need to be read for each token. Using indexes makes it easy to stop searching when there are no more sequences for a token, and the first sequence assigned is always very quickly available (it's the first element of each row). * mamba : support llama_kv_cache_seq_cp copy chains * mamba : support shifting and dividing the kv cache pos * mamba : make the server and parallel examples work with whole sequences A seq_id is dedicated to the system prompt in both cases. * llama : make llama_kv_cache_seq_rm return whether it succeeded or not * mamba : dedicate an input tensor for state copy indices This is cleaner and makes it easier to adapt when/if token positions (and by extension, inp_K_shift) are no longer integers. * mamba : adapt perplexity, batched, and batched-bench examples * perplexity : limit the max number of sequences This adapts to what the loaded model can provide. * llama : add llama_n_max_seq to get the upper limit for seq_ids Used by the perplexity example. * batched : pass n_parallel to the model's context params This should have been there already, but it wasn't. * batched-bench : reserve sequences to support Mamba * batched-bench : fix tokens being put in wrong sequences Generation quality isn't what's measured in there anyway, but at least using the correct sequences avoids using non-consecutive token positions. * mamba : stop abusing attention metadata This breaks existing converted-to-GGUF Mamba models, but will allow supporting mixed architectures like MambaFormer without needing to break Mamba models. This will also allow changing the size of Mamba's states without having to reconvert models in the future. (e.g. using something else than d_conv - 1 columns for the conv_states will not require breaking existing converted Mamba models again) * gguf-py : add new KV metadata key-value pairs for Mamba * llama : add new metadata key-value pairs for Mamba * llama : guard against divisions by zero when n_head is 0 * mamba : rename "unlimited" KV cache property to "recurrent" * mamba : more correctly update the "used" field of the KV cache * ggml : in ggml_ssm_scan, use a threshold for soft_plus This is how the official Mamba implementation does it, and it's also what torch.nn.Softplus does. * convert : for Mamba, fallback to internal NeoX tokenizer The resulting models are exactly the same as if the tokenizer.json and tokenizer_config.json of GPT-NeoX were there. * mamba : support state saving and restoring * ggml : implicitly pass src tensors through dst for Mamba-related ops * mamba : clarify some comments * server : fix cache_tokens not getting correctly resized Otherwise, when the "we have to evaluate at least 1 token" special case was triggered, an extra token was kept in cache_tokens even if it was removed from the KV cache. For Mamba, this caused useless prompt reprocessing when the previous request triggered the above case. * convert-hf : support new metadata keys for Mamba For the models available at https://huggingface.co/collections/state-spaces/transformers-compatible-mamba-65e7b40ab87e5297e45ae406 * mamba : rename metadata to be more similar to transformers library This breaks existing converted-to-GGUF models, but the metadata names are more "standard". * mamba : support mamba-*-hf models These models share their token_embd.weight with their output.weight * mamba : add missing spaces This is purely a formatting change. * convert-hf : omit output.weight when identical with token_embd.weight Only for Mamba for now, but it might be relevant for other models eventually. Most Mamba models actually share these two tensors, albeit implicitly. * readme : add Mamba to supported models, and add recent API changes * mamba : move state_seq and state_mask views outside layer loop A few tensors were also missing `struct` in front of `ggml_tensor`.
2024-03-08 22:31:00 +00:00
// v is contiguous for recurrent models
// TODO: use other tensors for state models than k and v
const size_t v_size = ggml_row_size(kv_self.v_l[il]->type, n_embd_v_gqa*kv_head);
ggml_backend_tensor_set(kv_self.v_l[il], inp, 0, v_size);
inp += v_size;
continue;
}
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
// v is not contiguous, copy row by row
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
const size_t v_row_size = ggml_row_size(kv_self.v_l[il]->type, kv_head);
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
const size_t v_row_stride = ggml_row_size(kv_self.v_l[il]->type, kv_self.size);
llama : refactor k-shift implementation + KV defragmentation (#5691) * llama : refactor k-shift implementation ggml-ci * llama : rename llama_kv_cache_seq_shift to llama_kv_cache_seq_add * llama : cont k-shift refactoring + normalize type names ggml-ci * minor : fix MPI builds * llama : reuse n_rot from the build context ggml-ci * llama : revert enum name changes from this PR ggml-ci * llama : update llama_rope_type * llama : add comment about rope values * llama : fix build * passkey : apply kv cache updates explicitly ggml-ci * llama : change name to llama_kv_cache_update() * llama : add llama_kv_cache_seq_pos_max() * passkey : fix llama_kv_cache_seq_pos_max() usage * llama : some llama_kv_cell simplifications * llama : add llama_kv_cache_compress (EXPERIMENTAL) * llama : add alternative KV cache merging (EXPERIMENTAL) * llama : add llama_kv_cache_defrag * llama : comments * llama : remove llama_kv_cache_compress will add in a separate PR ggml-ci * llama : defragment via non-overlapping moves * llama : ggml_graph based defrag implementation ggml-ci * llama : switch the loop order in build_defrag * llama : add comments
2024-02-25 20:12:24 +00:00
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
for (int ir = 0; ir < (int) n_embd_v_gqa; ++ir) {
llama : fix session save/load with quantized KV (#5649)
2024-02-21 21:52:39 +00:00
ggml_backend_tensor_set(kv_self.v_l[il], inp, ir*v_row_stride, v_row_size);
llama : ggml-backend integration (#4766) * llama : ggml-backend integration * ggml-backend : add names to buffers * fix unmap after loading * batched-bench : add tensor_split param * llama : check for null tensor_split * ggml-backend : increase GGML_MAX_BACKENDS * improve graph splitting, partial fix for --no-kv-offload * cuda : add ggml-backend split buffer support * cuda : do not create buffer types for devices that don't exist (fixes usage without CUDA devices available) * ggml : fix null backend dereference (#4807) * ggml : fix null backend dereference * ggml : also check ggml_backend_is_cpu * test-backend-ops : check buffer allocation failures * llama : add cparam (split_mode) and command line argument (--split-mode, -sm) to configure the split mode (none, layer or row) * ggml : fix mul_mat_id work size * llama : rewrite session kv load/set without graphs * minor * llama : only initialize used backends, free backends on context free * llama : abort ctx if cuda backend init fails * llama : rewrite lora with ggml-backend and compute on CPU ggml-ci * llama : only map to a backend buffer the region of the file mapping containing the tensors used in the buffer * opencl : add ggml-backend buffer type * cuda : only use batched_cublas with batched mat muls (fixes fp16 tg perf) * llama : on Metal, by default offload the full model ggml-ci * metal : page align the data ptr (#4854) * Apply suggestions from code review Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * cuda : fix split buffer free * address review comments * llama-bench : add split-mode parameter * fix whitespace * opencl : fix double initialization * server : add --split-mode parameter * use async copy and compute to improve multi-gpu performance ggml-ci * use async memcpys to copy the graph outputs to the CPU * fix opencl * use a host buffer for the cpu compute buffer for faster copies to the gpu --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2024-01-12 19:07:38 +00:00
inp += v_row_size;
}
llama : initial ggml-backend integration (#4520) * llama : initial ggml-backend integration * add ggml-metal * cuda backend can be used though ggml-backend with LLAMA_GGML_BACKEND_CUDA_TEST access all tensor data with ggml_backend_tensor_get/set * add ggml_backend_buffer_clear zero-init KV cache buffer * add ggml_backend_buffer_is_hos, used to avoid copies if possible when accesing tensor data * disable gpu backends with ngl 0 * more accurate mlock * unmap offloaded part of the model * use posix_fadvise64(.., POSIX_FADV_SEQUENTIAL) to improve performance with mmap * update quantize and lora * update session copy/set to use ggml-backend ggml-ci * use posix_fadvise instead of posix_fadvise64 * ggml_backend_alloc_ctx_tensors_from_buft : remove old print * llama_mmap::align_offset : use pointers instead of references for out parameters * restore progress_callback behavior * move final progress_callback call to load_all_data * cuda : fix fprintf format string (minor) * do not offload scales * llama_mmap : avoid unmapping the same fragments again in the destructor * remove unnecessary unmap * metal : add default log function that prints to stderr, cleanup code ggml-ci --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-21 20:07:46 +00:00
}
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
GGML_ASSERT(kv_buf_size == inp - src - pre_kv_buf_size);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
}
llama : fix session saving/loading (#3400) * llama : fix session saving/loading * llama : temp fix for clearing "future" tokens from the KV cache * llama : fix handling of "future" tokens when loading sessions * llama : fix comments for llama_kv_cache API
2023-10-03 18:04:01 +00:00
ctx->kv_self.head = kv_head;
llama : KV cache view API + better KV cache management (#4170) * llama : keep track of used KV cells + better KV cache management * llama : zero KV cache used upon clear ggml-ci * llama : allow exporting a view of the KV cache (#4180) * Allow exporting a view of the KV cache * Allow dumping the sequences per cell in common * Track max contiguous cells value and position as well * Fix max contiguous empty cells index calculation Make dump functions deal with lengths or sequences counts > 10 better * Fix off by one error in dump_kv_cache_view * Add doc comments for KV cache view functions Eliminate cell sequence struct; use llama_seq_id directly Minor cleanups * common : add -dkvc arg for enabling kv cache dumps --------- Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-11-23 17:07:56 +00:00
ctx->kv_self.used = kv_used;
llama : fix session saving/loading (#3400) * llama : fix session saving/loading * llama : temp fix for clearing "future" tokens from the KV cache * llama : fix handling of "future" tokens when loading sessions * llama : fix comments for llama_kv_cache API
2023-10-03 18:04:01 +00:00
llama : fix llama_copy_state_data with fragmented KV cache (#5840) The row size of the saved states was based on kv_self.head while it should be based on llama_kv_cache_cell_max. Existing session files should still work. * llama : fix llama_kv_cache_cell_max inability to return 1 I've also changed its return type to uint32_t, because this function is always used to set the value of uint32_t variables, and because the index already has this type. * llama : fix state size calculation Some bytes in the state were unaccounted for in llama_get_state_size. Since the logits reserve so much space, it did not cause problems.
2024-03-03 08:41:55 +00:00
for (uint32_t i = 0; i < kv_head; ++i) {
llama : fix session saving/loading (#3400) * llama : fix session saving/loading * llama : temp fix for clearing "future" tokens from the KV cache * llama : fix handling of "future" tokens when loading sessions * llama : fix comments for llama_kv_cache API
2023-10-03 18:04:01 +00:00
llama_pos pos;
size_t seq_id_size;
memcpy(&pos, inp, sizeof(pos)); inp += sizeof(pos);
memcpy(&seq_id_size, inp, sizeof(seq_id_size)); inp += sizeof(seq_id_size);
ctx->kv_self.cells[i].pos = pos;
llama_seq_id seq_id;
for (size_t j = 0; j < seq_id_size; ++j) {
memcpy(&seq_id, inp, sizeof(seq_id)); inp += sizeof(seq_id);
ctx->kv_self.cells[i].seq_id.insert(seq_id);
}
}
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
}
llama : free ggml context in set / copy state data (close #1425)
2023-05-13 06:08:52 +00:00
const size_t nread = inp - src;
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
const size_t max_size = llama_state_get_size(ctx);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
GGML_ASSERT(nread <= max_size);
llama : refactor get / set state + remove redundant kv cache API (#1143)
2023-04-24 04:40:02 +00:00
return nread;
Added api for getting/setting the kv_cache (#685) The api provides access methods for retrieving the current memory buffer for the kv_cache and its token number. It also contains a method for setting the kv_cache from a memory buffer. This makes it possible to load/save history - maybe support --cache-prompt paramater as well? Co-authored-by: Pavol Rusnak <pavol@rusnak.io>
2023-04-02 10:23:04 +00:00
}
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
static bool llama_state_load_file_internal(struct llama_context * ctx, const char * path_session, llama_token * tokens_out, size_t n_token_capacity, size_t * n_token_count_out) {
llama : fix session load / save (#1263)
2023-05-01 11:54:59 +00:00
llama_file file(path_session, "rb");
// sanity checks
{
const uint32_t magic = file.read_u32();
const uint32_t version = file.read_u32();
llama : fix various warnings
2023-05-13 08:23:15 +00:00
if (magic != LLAMA_SESSION_MAGIC || version != LLAMA_SESSION_VERSION) {
add log_callback to llama_context_params for custom logging. (#2234) * add log_callback to llama_context_params for custom logging. * Fix macro expansion on gcc * Add struct llama_state for global variables and move log_callback there * Turn log level into enum and some minor changes. * Remove model_for_logging parameter (not needed anymore) * Convert remaining fprintf(stderr, ...) calls to use new macros. * Fix enum and initialize g_state * Fix log calls after merge * Fix missing static * Add back all the new lines in the logging strings * Add comment for llama_log_callback and replace remaining printf calls --------- Co-authored-by: grahameth <-> Co-authored-by: Helmut <helmut.buhler@inf.h-brs.de>
2023-08-09 20:46:40 +00:00
LLAMA_LOG_ERROR("%s : unknown (magic, version) for session file: %08x, %08x\n", __func__, magic, version);
llama : fix session load / save (#1263)
2023-05-01 11:54:59 +00:00
return false;
}
llama_hparams session_hparams;
file.read_raw(&session_hparams, sizeof(llama_hparams));
if (session_hparams != ctx->model.hparams) {
add log_callback to llama_context_params for custom logging. (#2234) * add log_callback to llama_context_params for custom logging. * Fix macro expansion on gcc * Add struct llama_state for global variables and move log_callback there * Turn log level into enum and some minor changes. * Remove model_for_logging parameter (not needed anymore) * Convert remaining fprintf(stderr, ...) calls to use new macros. * Fix enum and initialize g_state * Fix log calls after merge * Fix missing static * Add back all the new lines in the logging strings * Add comment for llama_log_callback and replace remaining printf calls --------- Co-authored-by: grahameth <-> Co-authored-by: Helmut <helmut.buhler@inf.h-brs.de>
2023-08-09 20:46:40 +00:00
LLAMA_LOG_INFO("%s : model hparams didn't match from session file!\n", __func__);
llama : fix session load / save (#1263)
2023-05-01 11:54:59 +00:00
return false;
}
}
// load the prompt
{
const uint32_t n_token_count = file.read_u32();
if (n_token_count > n_token_capacity) {
add log_callback to llama_context_params for custom logging. (#2234) * add log_callback to llama_context_params for custom logging. * Fix macro expansion on gcc * Add struct llama_state for global variables and move log_callback there * Turn log level into enum and some minor changes. * Remove model_for_logging parameter (not needed anymore) * Convert remaining fprintf(stderr, ...) calls to use new macros. * Fix enum and initialize g_state * Fix log calls after merge * Fix missing static * Add back all the new lines in the logging strings * Add comment for llama_log_callback and replace remaining printf calls --------- Co-authored-by: grahameth <-> Co-authored-by: Helmut <helmut.buhler@inf.h-brs.de>
2023-08-09 20:46:40 +00:00
LLAMA_LOG_ERROR("%s : token count in session file exceeded capacity! %u > %zu\n", __func__, n_token_count, n_token_capacity);
llama : fix session load / save (#1263)
2023-05-01 11:54:59 +00:00
return false;
}
file.read_raw(tokens_out, sizeof(llama_token) * n_token_count);
*n_token_count_out = n_token_count;
}
// restore the context state
{
const size_t n_state_size_cur = file.size - file.tell();
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
const size_t n_state_size_max = llama_state_get_size(ctx);
llama : fix session load / save (#1263)
2023-05-01 11:54:59 +00:00
llama : only copy used KV cache in get / set state (#1272) * llama : only copy used KV cache in get / set state * switch to ggml for copying k, v * avoid designated initializers
2023-05-03 02:26:13 +00:00
if (n_state_size_cur > n_state_size_max) {
add log_callback to llama_context_params for custom logging. (#2234) * add log_callback to llama_context_params for custom logging. * Fix macro expansion on gcc * Add struct llama_state for global variables and move log_callback there * Turn log level into enum and some minor changes. * Remove model_for_logging parameter (not needed anymore) * Convert remaining fprintf(stderr, ...) calls to use new macros. * Fix enum and initialize g_state * Fix log calls after merge * Fix missing static * Add back all the new lines in the logging strings * Add comment for llama_log_callback and replace remaining printf calls --------- Co-authored-by: grahameth <-> Co-authored-by: Helmut <helmut.buhler@inf.h-brs.de>
2023-08-09 20:46:40 +00:00
LLAMA_LOG_ERROR("%s : the state size in session file is too big! max %zu, got %zu\n", __func__, n_state_size_max, n_state_size_cur);
llama : fix session load / save (#1263)
2023-05-01 11:54:59 +00:00
return false;
}
llama : only copy used KV cache in get / set state (#1272) * llama : only copy used KV cache in get / set state * switch to ggml for copying k, v * avoid designated initializers
2023-05-03 02:26:13 +00:00
std::vector<uint8_t> state_data(n_state_size_max);
llama : fix session load / save (#1263)
2023-05-01 11:54:59 +00:00
file.read_raw(state_data.data(), n_state_size_cur);
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
llama_state_set_data(ctx, state_data.data());
llama : fix session load / save (#1263)
2023-05-01 11:54:59 +00:00
}
llama : fix return value of llama_load_session_file_internal (#2022)
2023-07-01 16:05:09 +00:00
return true;
llama : catch llama_load_session_file_internal exceptions (#2022) * convert checks in llama_load_session_file to throw and handle them * make llama_load_session_file_internal static * address feedbacks to avoid using exceptions
2023-07-01 16:02:58 +00:00
}
llama : fix session load / save (#1263)
2023-05-01 11:54:59 +00:00
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
bool llama_state_load_file(struct llama_context * ctx, const char * path_session, llama_token * tokens_out, size_t n_token_capacity, size_t * n_token_count_out) {
llama : catch llama_load_session_file_internal exceptions (#2022) * convert checks in llama_load_session_file to throw and handle them * make llama_load_session_file_internal static * address feedbacks to avoid using exceptions
2023-07-01 16:02:58 +00:00
try {
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
return llama_state_load_file_internal(ctx, path_session, tokens_out, n_token_capacity, n_token_count_out);
llama : catch llama_load_session_file_internal exceptions (#2022) * convert checks in llama_load_session_file to throw and handle them * make llama_load_session_file_internal static * address feedbacks to avoid using exceptions
2023-07-01 16:02:58 +00:00
} catch (const std::exception & err) {
add log_callback to llama_context_params for custom logging. (#2234) * add log_callback to llama_context_params for custom logging. * Fix macro expansion on gcc * Add struct llama_state for global variables and move log_callback there * Turn log level into enum and some minor changes. * Remove model_for_logging parameter (not needed anymore) * Convert remaining fprintf(stderr, ...) calls to use new macros. * Fix enum and initialize g_state * Fix log calls after merge * Fix missing static * Add back all the new lines in the logging strings * Add comment for llama_log_callback and replace remaining printf calls --------- Co-authored-by: grahameth <-> Co-authored-by: Helmut <helmut.buhler@inf.h-brs.de>
2023-08-09 20:46:40 +00:00
LLAMA_LOG_ERROR("error loading session file: %s\n", err.what());
llama : catch llama_load_session_file_internal exceptions (#2022) * convert checks in llama_load_session_file to throw and handle them * make llama_load_session_file_internal static * address feedbacks to avoid using exceptions
2023-07-01 16:02:58 +00:00
return false;
}
llama : fix session load / save (#1263)
2023-05-01 11:54:59 +00:00
}
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
static bool llama_state_save_file_internal(struct llama_context * ctx, const char * path_session, const llama_token * tokens, size_t n_token_count) {
llama : fix session load / save (#1263)
2023-05-01 11:54:59 +00:00
llama_file file(path_session, "wb");
file.write_u32(LLAMA_SESSION_MAGIC);
file.write_u32(LLAMA_SESSION_VERSION);
file.write_raw(&ctx->model.hparams, sizeof(llama_hparams));
// save the prompt
file.write_u32((uint32_t) n_token_count);
file.write_raw(tokens, sizeof(llama_token) * n_token_count);
Stream save llama context data to file instead of allocating entire buffer upfront (#2488) * added stream saving context data to file to avoid allocating unnecessary amounts of memory * generalised copying state data to file or buffer * added comments explaining how copy_state_data works * fixed trailing whitespaces * fixed save load state example * updated save load state to use public function in llama.cpp * - restored breakage of the llama_copy_state_data API - moved new logic for copying llama state data to internal function * fixed function declaration order * restored save load state example * fixed whitepace * removed unused llama-util.h include * Apply suggestions from code review Co-authored-by: slaren <slarengh@gmail.com> * Apply code review suggestions Co-authored-by: slaren <slarengh@gmail.com> --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-08-04 11:29:52 +00:00
// save the context state using stream saving
llama_data_file_context data_ctx(&file);
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
llama_state_get_data_internal(ctx, &data_ctx);
llama : fix session load / save (#1263)
2023-05-01 11:54:59 +00:00
return true;
}
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
bool llama_state_save_file(struct llama_context * ctx, const char * path_session, const llama_token * tokens, size_t n_token_count) {
try {
return llama_state_save_file_internal(ctx, path_session, tokens, n_token_count);
} catch (const std::exception & err) {
LLAMA_LOG_ERROR("error saving session file: %s\n", err.what());
return false;
}
}
size_t llama_state_seq_get_size(struct llama_context* ctx, llama_seq_id seq_id) {
// save the size of size_t as a uint32_t for safety check
const size_t size_t_size_size = sizeof(uint32_t);
// other values
const size_t s_cell_count_size = sizeof(uint32_t);
const size_t s_layer_count_size = sizeof(uint32_t);
const size_t n_embd_v_gqa_size = sizeof(uint32_t);
size_t s_cell_count = 0;
size_t s_cell_data_size = 0;
const auto & kv_self = ctx->kv_self;
const auto & hparams = ctx->model.hparams;
const uint32_t n_layer = hparams.n_layer;
const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa() + hparams.n_embd_k_s();
const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa() + hparams.n_embd_v_s();
for (uint32_t i = 0; i < kv_self.size; ++i) {
const auto & cell = kv_self.cells[i];
if (cell.seq_id.count(seq_id) > 0) {
++s_cell_count;
s_cell_data_size += sizeof(llama_pos);
}
}
for (int il = 0; il < (int)n_layer; ++il) {
// types of keys and values
s_cell_data_size += sizeof(int32_t) * 2;
// k_size_row and v_size_el values of layer
s_cell_data_size += sizeof(size_t) * 2;
// keys
const size_t k_size_row = ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa);
s_cell_data_size += k_size_row * s_cell_count;
// values (transposed)
const size_t v_size_el = ggml_type_size(kv_self.v_l[il]->type);
s_cell_data_size += v_size_el * s_cell_count * n_embd_v_gqa;
}
const size_t s_total = (
size_t_size_size +
s_cell_count_size +
s_layer_count_size +
n_embd_v_gqa_size +
s_cell_data_size
);
return s_total;
}
static size_t llama_state_seq_get_data_internal(struct llama_context * ctx, llama_data_context & data_ctx, llama_seq_id seq_id) {
llama : synchronize before get/set session data (#6911)
2024-04-25 15:59:03 +00:00
llama_synchronize(ctx);
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
const auto & kv_self = ctx->kv_self;
GGML_ASSERT(!kv_self.recurrent); // not implemented
// Save the size of size_t as a uint32_t for safety check
const uint32_t size_t_size = sizeof(size_t);
data_ctx.write(&size_t_size, sizeof(size_t_size));
std::vector<std::pair<uint32_t, uint32_t>> cell_ranges; // ranges, from inclusive, to exclusive
uint32_t cell_count = 0;
// Count the number of cells with the specified seq_id
// Find all the ranges of cells with this seq id
{
uint32_t cell_range_begin = kv_self.size;
for (uint32_t i = 0; i < kv_self.size; ++i) {
const auto & cell = kv_self.cells[i];
if (cell.has_seq_id(seq_id)) {
++cell_count;
if (cell_range_begin == kv_self.size) {
cell_range_begin = i;
}
}
else {
if (cell_range_begin != kv_self.size) {
cell_ranges.push_back({ cell_range_begin, i });
cell_range_begin = kv_self.size;
}
}
}
if (cell_range_begin != kv_self.size) {
cell_ranges.push_back({ cell_range_begin, kv_self.size });
}
// DEBUG CHECK: Sum of cell counts in ranges should equal the total cell count
uint32_t cell_count_check = 0;
for (const auto & range : cell_ranges) {
cell_count_check += range.second - range.first;
}
GGML_ASSERT(cell_count == cell_count_check);
}
// Write the cell count
data_ctx.write(&cell_count, sizeof(cell_count));
const auto & hparams = ctx->model.hparams;
const uint32_t n_layer = hparams.n_layer;
const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa() + hparams.n_embd_k_s();
const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa() + hparams.n_embd_v_s();
// Write the layer count
data_ctx.write(&n_layer, sizeof(n_layer));
// Write n_embd_v_gqa
data_ctx.write(&n_embd_v_gqa, sizeof(n_embd_v_gqa));
// Iterate the ranges and write all the pos (this is the token position in the prompt)
for (const auto & range : cell_ranges) {
for (uint32_t i = range.first; i < range.second; ++i) {
const auto & cell = kv_self.cells[i];
data_ctx.write(&cell.pos, sizeof(cell.pos));
}
}
// Iterate and write all the keys first, each row is a cell
// Get whole range at a time
std::vector<uint8_t> tmp_buf;
for (int il = 0; il < (int)n_layer; ++il) {
// Write key type
const int32_t k_type_i = (int32_t)kv_self.k_l[il]->type;
data_ctx.write(&k_type_i, sizeof(k_type_i));
// Write row size of key
const size_t k_size_row = ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa);
data_ctx.write(&k_size_row, sizeof(k_size_row));
// Read each range of cells of k_size length each into tmp_buf and write out
for (const auto & range : cell_ranges) {
const size_t range_size = range.second - range.first;
tmp_buf.resize(range_size * k_size_row);
ggml_backend_tensor_get(kv_self.k_l[il], tmp_buf.data(), range.first * k_size_row, range_size * k_size_row);
data_ctx.write(tmp_buf.data(), tmp_buf.size());
}
}
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
// TODO: simplify, reduce copy-paste
if (!kv_self.v_trans) {
for (int il = 0; il < (int)n_layer; ++il) {
// Write value type
const int32_t v_type_i = (int32_t)kv_self.v_l[il]->type;
data_ctx.write(&v_type_i, sizeof(v_type_i));
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
// Write row size of value
const size_t v_size_row = ggml_row_size(kv_self.v_l[il]->type, n_embd_v_gqa);
data_ctx.write(&v_size_row, sizeof(v_size_row));
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
// Read each range of cells of v_size length each into tmp_buf and write out
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
for (const auto & range : cell_ranges) {
const size_t range_size = range.second - range.first;
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
tmp_buf.resize(range_size * v_size_row);
ggml_backend_tensor_get(kv_self.v_l[il], tmp_buf.data(), range.first * v_size_row, range_size * v_size_row);
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
data_ctx.write(tmp_buf.data(), tmp_buf.size());
}
}
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
} else {
// For the values, they are transposed, so we also need the element size and get the element ranges from each row
const uint32_t kv_size = kv_self.size;
for (int il = 0; il < (int)n_layer; ++il) {
// Write value type
const int32_t v_type_i = (int32_t)kv_self.v_l[il]->type;
data_ctx.write(&v_type_i, sizeof(v_type_i));
// Write element size
const size_t v_size_el = ggml_type_size(kv_self.v_l[il]->type);
data_ctx.write(&v_size_el, sizeof(v_size_el));
// For each row, we get the element values of each cell
for (uint32_t j = 0; j < n_embd_v_gqa; ++j) {
// Read each range of cells of v_size_el length each into tmp_buf and write out
for (const auto & range : cell_ranges) {
const size_t range_size = range.second - range.first;
const size_t src_offset = (range.first + j * kv_size) * v_size_el;
tmp_buf.resize(range_size * v_size_el);
ggml_backend_tensor_get(kv_self.v_l[il], tmp_buf.data(), src_offset, tmp_buf.size());
data_ctx.write(tmp_buf.data(), tmp_buf.size());
}
}
}
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
}
return data_ctx.get_size_written();
}
size_t llama_state_seq_get_data(struct llama_context* ctx, uint8_t* dst, llama_seq_id seq_id) {
llama_data_buffer_context data_ctx(dst);
return llama_state_seq_get_data_internal(ctx, data_ctx, seq_id);
}
size_t llama_state_seq_set_data(struct llama_context * ctx, const uint8_t * src, llama_seq_id dest_seq_id) {
llama : synchronize before get/set session data (#6911)
2024-04-25 15:59:03 +00:00
llama_synchronize(ctx);
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
auto & kv_self = ctx->kv_self;
GGML_ASSERT(!kv_self.recurrent); // not implemented
// Wipe the slot
llama_kv_cache_seq_rm(kv_self, dest_seq_id, -1, -1);
const uint8_t * inp = src;
// Read size of size_t
uint32_t size_t_size;
memcpy(&size_t_size, inp, sizeof(size_t_size));
inp += sizeof(size_t_size);
if (size_t_size != sizeof(size_t)) {
LLAMA_LOG_ERROR("%s: size_t size mismatch\n", __func__);
return 0;
}
// Read the cell count
uint32_t cell_count;
memcpy(&cell_count, inp, sizeof(cell_count));
inp += sizeof(cell_count);
// Read the layer count
uint32_t n_layer_ref;
memcpy(&n_layer_ref, inp, sizeof(n_layer_ref));
inp += sizeof(n_layer_ref);
// Read n_embd_v_gqa
uint32_t n_embd_v_gqa_ref;
memcpy(&n_embd_v_gqa_ref, inp, sizeof(n_embd_v_gqa_ref));
inp += sizeof(n_embd_v_gqa_ref);
// Sanity check model compatibility
const auto & hparams = ctx->model.hparams;
const uint32_t n_layer = hparams.n_layer;
const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa() + hparams.n_embd_k_s();
const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa() + hparams.n_embd_v_s();
if (n_layer != n_layer_ref) {
LLAMA_LOG_ERROR("%s: mismatched n_layer (%d != %d)\n", __func__, n_layer, n_layer_ref);
return 0;
}
if (n_embd_v_gqa != n_embd_v_gqa_ref) {
LLAMA_LOG_ERROR("%s: mismatched n_embd_v_gqa (%d != %d)\n", __func__, n_embd_v_gqa, n_embd_v_gqa_ref);
return 0;
}
// Allocate the new cells for the slot
if (cell_count) {
llama_batch batch = llama_batch_init(cell_count, 0, 1);
batch.n_tokens = cell_count;
for (uint32_t i = 0; i < cell_count; ++i) {
llama_pos pos;
memcpy(&pos, inp, sizeof(pos));
inp += sizeof(pos);
batch.pos[i] = pos;
batch.n_seq_id[i] = 1;
batch.seq_id[i][0] = dest_seq_id;
}
if (!llama_kv_cache_find_slot(kv_self, batch)) {
llama_batch_free(batch);
LLAMA_LOG_ERROR("%s: failed to find available cells in kv cache\n", __func__);
return 0;
}
// DEBUG CHECK: kv_self.head should be our first cell, kv_self.head + cell_count - 1 should be our last cell (verify seq_id and pos values)
// Assume that this is one contiguous block of cells
GGML_ASSERT(kv_self.head + cell_count <= kv_self.size);
GGML_ASSERT(kv_self.cells[kv_self.head].pos == batch.pos[0]);
GGML_ASSERT(kv_self.cells[kv_self.head + cell_count - 1].pos == batch.pos[cell_count - 1]);
GGML_ASSERT(kv_self.cells[kv_self.head].has_seq_id(dest_seq_id));
GGML_ASSERT(kv_self.cells[kv_self.head + cell_count - 1].has_seq_id(dest_seq_id));
// Cleanup
llama_batch_free(batch);
}
const uint32_t kv_size = kv_self.size;
const uint32_t kv_head = kv_self.head;
// For each layer, read the keys for each cell, one row is one cell, read as one contiguous blo
for (int il = 0; il < (int)n_layer; ++il) {
// Read type of key
int32_t k_type_i_ref;
memcpy(&k_type_i_ref, inp, sizeof(k_type_i_ref));
inp += sizeof(k_type_i_ref);
const int32_t k_type_i = (int32_t)kv_self.k_l[il]->type;
if (k_type_i != k_type_i_ref) {
llama_kv_cache_seq_rm(kv_self, dest_seq_id, -1, -1);
LLAMA_LOG_ERROR("%s: mismatched key type (%d != %d, layer %d)\n", __func__, k_type_i, k_type_i_ref, il);
return 0;
}
// Read row size of key
size_t k_size_row_ref;
memcpy(&k_size_row_ref, inp, sizeof(k_size_row_ref));
inp += sizeof(k_size_row_ref);
const size_t k_size_row = ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa);
if (k_size_row != k_size_row_ref) {
llama_kv_cache_seq_rm(kv_self, dest_seq_id, -1, -1);
LLAMA_LOG_ERROR("%s: mismatched key row size (%zu != %zu, layer %d)\n", __func__, k_size_row, k_size_row_ref, il);
return 0;
}
if (cell_count) {
// Read and set the keys for the whole cell range
ggml_backend_tensor_set(kv_self.k_l[il], inp, kv_head * k_size_row, cell_count * k_size_row);
inp += cell_count * k_size_row;
}
}
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
// TODO: simplify, reduce copy-paste
if (!kv_self.v_trans) {
for (int il = 0; il < (int)n_layer; ++il) {
// Read type of value
int32_t v_type_i_ref;
memcpy(&v_type_i_ref, inp, sizeof(v_type_i_ref));
inp += sizeof(v_type_i_ref);
const int32_t v_type_i = (int32_t)kv_self.v_l[il]->type;
if (v_type_i != v_type_i_ref) {
llama_kv_cache_seq_rm(kv_self, dest_seq_id, -1, -1);
LLAMA_LOG_ERROR("%s: mismatched value type (%d != %d, layer %d)\n", __func__, v_type_i, v_type_i_ref, il);
return 0;
}
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
// Read row size of value
size_t v_size_row_ref;
memcpy(&v_size_row_ref, inp, sizeof(v_size_row_ref));
inp += sizeof(v_size_row_ref);
const size_t v_size_row = ggml_row_size(kv_self.v_l[il]->type, n_embd_v_gqa);
if (v_size_row != v_size_row_ref) {
llama_kv_cache_seq_rm(kv_self, dest_seq_id, -1, -1);
LLAMA_LOG_ERROR("%s: mismatched value row size (%zu != %zu, layer %d)\n", __func__, v_size_row, v_size_row_ref, il);
return 0;
}
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
if (cell_count) {
// Read and set the values for the whole cell range
ggml_backend_tensor_set(kv_self.v_l[il], inp, kv_head * v_size_row, cell_count * v_size_row);
inp += cell_count * v_size_row;
}
}
} else {
// For each layer, read the values for each cell (transposed)
for (int il = 0; il < (int)n_layer; ++il) {
// Read type of value
int32_t v_type_i_ref;
memcpy(&v_type_i_ref, inp, sizeof(v_type_i_ref));
inp += sizeof(v_type_i_ref);
const int32_t v_type_i = (int32_t)kv_self.v_l[il]->type;
if (v_type_i != v_type_i_ref) {
llama_kv_cache_seq_rm(kv_self, dest_seq_id, -1, -1);
LLAMA_LOG_ERROR("%s: mismatched value type (%d != %d, layer %d)\n", __func__, v_type_i, v_type_i_ref, il);
return 0;
}
// Read element size of value
size_t v_size_el_ref;
memcpy(&v_size_el_ref, inp, sizeof(v_size_el_ref));
inp += sizeof(v_size_el_ref);
const size_t v_size_el = ggml_type_size(kv_self.v_l[il]->type);
if (v_size_el != v_size_el_ref) {
llama_kv_cache_seq_rm(kv_self, dest_seq_id, -1, -1);
LLAMA_LOG_ERROR("%s: mismatched value element size (%zu != %zu, layer %d)\n", __func__, v_size_el, v_size_el_ref, il);
return 0;
}
if (cell_count) {
// For each row in the transposed matrix, read the values for the whole cell range
for (uint32_t j = 0; j < n_embd_v_gqa; ++j) {
const size_t dst_offset = (kv_head + j * kv_size) * v_size_el;
ggml_backend_tensor_set(kv_self.v_l[il], inp, dst_offset, cell_count * v_size_el);
inp += cell_count * v_size_el;
}
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
}
}
}
const size_t nread = inp - src;
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 09:16:08 +00:00
llama : save and restore kv cache for single seq id (#6341) * llama : save and restore kv cache for single seq id * remove trailing whitespace * respond error in case there's no space in the kv cache * add kv seq save restore to test case * add --slot-save-path arg to enable save restore and restrict save location * Returning 0 for some cases, instead of asserting. * cleanup error cases * rename sequence state functions * rename state get set functions * add previous function names back in with DEPRECATED notice * update doc * adjust endpoints to preferred style * fix restoring zero cell count * handle seq rm return value * unused param * keep in the size check * fix return types * add server test case for slot save restore * cleanup * add cake * cleanup style * add special * removing a whole sequence never fails * move sequence state file functionality from server to llama to match session api and add version tags * catch exceptions on save as well * error log messages * check types for stricter restore * update server doc * readme : update API changes date * strict filename validation * move include, reject bom as well * also reject empty filename * reject whitespace and trailing dot --------- Co-authored-by: Martin Evans <martindevans@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-08 12:43:30 +00:00
return nread;
}
static size_t llama_state_seq_save_file_internal(struct llama_context * ctx, const char * filepath, llama_seq_id seq_id, const llama_token * tokens, size_t n_token_count) {
llama_file file(filepath, "wb");
file.write_u32(LLAMA_STATE_SEQ_MAGIC);
file.write_u32(LLAMA_STATE_SEQ_VERSION);
// save the prompt
file.write_u32((uint32_t)n_token_count);
file.write_raw(tokens, sizeof(llama_token) * n_token_count);
// save the context state using stream saving
llama_data_file_context data_ctx(&file);
llama_state_seq_get_data_internal(ctx, data_ctx, seq_id);
const size_t res = file.tell();
GGML_ASSERT(res == sizeof(uint32_t) * 3 + sizeof(llama_token) * n_token_count + data_ctx.get_size_written());
return res;
}
static size_t llama_state_seq_load_file_internal(struct llama_context * ctx, const char * filepath, llama_seq_id dest_seq_id, llama_token * tokens_out, size_t n_token_capacity, size_t * n_token_count_out) {
llama_file file(filepath, "rb");
// version checks
{
const uint32_t magic = file.read_u32();
const uint32_t version = file.read_u32();
if (magic != LLAMA_STATE_SEQ_MAGIC || version != LLAMA_STATE_SEQ_VERSION) {
LLAMA_LOG_ERROR("%s: unknown (magic, version) for sequence state file: %08x, %08x\n", __func__, magic, version);
return 0;
}
}
// load the prompt
{
const uint32_t n_token_count = file.read_u32();
if (n_token_count > n_token_capacity) {
LLAMA_LOG_ERROR("%s: token count in sequence state file exceeded capacity! %u > %zu\n", __func__, n_token_count, n_token_capacity);
return 0;
}
file.read_raw(tokens_out, sizeof(llama_token) * n_token_count);
*n_token_count_out = n_token_count;
}
// restore the context state
{
const size_t state_size = file.size - file.tell();
std::vector<uint8_t> state_data(state_size);
file.read_raw(state_data.data(), state_size);
const size_t nread = llama_state_seq_set_data(ctx, state_data.data(), dest_seq_id);
if (!nread) {
LLAMA_LOG_ERROR("%s: failed to restore sequence state\n", __func__);
return 0;
}
GGML_ASSERT(nread <= state_size);
GGML_ASSERT(nread + sizeof(uint32_t) * 3 + sizeof(llama_token) * *n_token_count_out == file.tell());
}
return file.tell();
}
size_t llama_state_seq_save_file(struct llama_context * ctx, const char * filepath, llama_seq_id seq_id, const llama_token * tokens, size_t n_token_count) {
try {
return llama_state_seq_save_file_internal(ctx, filepath, seq_id, tokens, n_token_count);
} catch (const std::exception & err) {
LLAMA_LOG_ERROR("error saving sequence state file: %s\n", err.what());
return 0;
}
}
size_t llama_state_seq_load_file(struct llama_context * ctx, const char * filepath, llama_seq_id dest_seq_id, llama_token * tokens_out, size_t n_token_capacity, size_t * n_token_count_out) {
try {
return llama_state_seq_load_file_internal(ctx, filepath, dest_seq_id, tokens_out, n_token_capacity, n_token_count_out);
} catch (const std::exception & err) {
LLAMA_LOG_ERROR("error loading sequence state file: %s\n", err.what());
return 0;
}
}
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
void llama_set_n_threads(struct llama_context * ctx, uint32_t n_threads, uint32_t n_threads_batch) {
ctx->cparams.n_threads = n_threads;
ctx->cparams.n_threads_batch = n_threads_batch;
}
llama : add abort_callback to interrupt computation (#5409) * using abort_callback from ggml to stop llama computation * format fix * a brief explaining comment --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-02 19:52:25 +00:00
void llama_set_abort_callback(struct llama_context * ctx, bool (*abort_callback)(void * data), void * abort_callback_data) {
ctx->abort_callback = abort_callback;
ctx->abort_callback_data = abort_callback_data;
}
llama : add support for GritLM (#5959) * add gritlm example * gritlm results match * tabs to spaces * comment out debug printing * rebase to new embed * gritlm embeddings are back babeee * add to gitignore * allow to toggle embedding mode * Clean-up GritLM sample code. * Fix types. * Flush stdout and output ending newline if streaming. * mostly style fixes; correct KQ_mask comment * add causal_attn flag to llama_cparams * gritml : minor * llama : minor --------- Co-authored-by: Douglas Hanley <thesecretaryofwar@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-10 15:56:30 +00:00
void llama_set_causal_attn(struct llama_context * ctx, bool causal_attn) {
ctx->cparams.causal_attn = causal_attn;
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
struct llama_batch llama_batch_get_one(
llama_token * tokens,
int32_t n_tokens,
llama_pos pos_0,
llama_seq_id seq_id) {
return {
speculative : add tree-based sampling example (#3624) * sampling : one sequence per sampling context ggml-ci * speculative : add tree-based sampling support ggml-ci * speculative : reuse the n_parallel CLI param * speculative : refactor sampling * examples : fix build after sampling refactoring ggml-ci * batched : fix n_seq_id * sampling : fix malloc ggml-ci * swift : fix build ggml-ci * swift : try to fix build ggml-ci * prompts : add assistant.txt * common : add llama_batch_add() and llama_batch_clear() helpers * speculative : minor refactor ggml-ci * minor : comments + rename ggml-ci * speculative : fix off-by-one for n_drafted * speculative : fix the n_drafted fix + p constants
2023-10-18 13:21:57 +00:00
/*n_tokens =*/ n_tokens,
/*tokens =*/ tokens,
/*embd =*/ nullptr,
/*pos =*/ nullptr,
/*n_seq_id =*/ nullptr,
/*seq_id =*/ nullptr,
/*logits =*/ nullptr,
/*all_pos_0 =*/ pos_0,
/*all_pos_1 =*/ 1,
/*all_seq_id =*/ seq_id,
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
};
}
llama : Metal inference (#1642) * mtl : export the LLaMA computation graph * ci : disable temporary * mtl : adapt the MNIST example as starter * mtl : no need for mtl-export tool, add cli arg for main instead * mtl : export just a small part of the graph for now to make it easier * mtl : move MSL code into separate file for easy editing * mtl : initial get_rows_q4_0 kernel * mtl : confirmed get_rows_q4_0 is working correctly * mtl : add rms_norm kernel + confirm working * mtl : add mul kernel + confirm working * mtl : initial mul_mat Q4 kernel (wrong results) * mtl : mul_mat fixes (still wrong) * mtl : another mul_mat Q4 (still does not work) * mtl : working mul_mat q4 * ggml : fix handling of "view" ops in ggml_graph_import() * mtl : add rope kernel * mtl : add reshape and transpose handling * ggml : store offset as opt arg for ggml_view_xd() operators * mtl : add cpy kernel + handle view ops * mtl : confirm f16 x f32 attention mul mat * mtl : add scale kernel * mtl : add diag_mask_inf kernel * mtl : fix soft_max kernel * ggml : update ggml_nbytes() to handle non-contiguous tensors * mtl : verify V tensor contents * mtl : add f32 -> f32 cpy kernel * mtl : add silu kernel * mtl : add non-broadcast mul kernel * mtl : full GPU inference of the computation graph * mtl : optimize rms_norm and soft_max kernels * mtl : add f16 mat x f32 vec multiplication kernel * mtl : fix bug in f16 x f32 mul mat + speed-up computation * mtl : faster mul_mat_q4_0_f32 kernel * mtl : fix kernel signature + roll inner loop * mtl : more threads for rms_norm + better timing * mtl : remove printfs from inner loop * mtl : simplify implementation * mtl : add save/load vocab to ggml file * mtl : plug Metal inference into llama.cpp (very quick-n-dirty) * mtl : make it work with main example Lots of hacks but at least now it generates text * mtl : preparing for merge * mtl : clean-up ggml mtl interface + suport scratch / inplace * mtl : remove temp / debug code * metal : final refactoring and simplification * Revert "ci : disable temporary" This reverts commit 98c267fc77fe811082f672538fc91bcfc9072d63. * metal : add comments * metal : clean-up stuff, fix typos * readme : add Metal instructions * readme : add example for main
2023-06-04 20:34:30 +00:00
llama : fix memory leak in llama_batch_free (#5252) The llama_batch_init allocates memory for a fixed number of tokens. However, the llama_batch_free only frees memory for the number of tokens that were added to the batch. This change-set uses a null terminated array for the batch seq_id, and frees all the elements until the nullptr is reached. This change-set also changes the name of the first parameter from `n_tokens` to `n_tokens_alloc` to more clearly indicate that this value is the number of tokens allocated to the batch, not the number of tokens in the batch.
2024-02-02 07:20:13 +00:00
struct llama_batch llama_batch_init(int32_t n_tokens_alloc, int32_t embd, int32_t n_seq_max) {
speculative : add tree-based sampling example (#3624) * sampling : one sequence per sampling context ggml-ci * speculative : add tree-based sampling support ggml-ci * speculative : reuse the n_parallel CLI param * speculative : refactor sampling * examples : fix build after sampling refactoring ggml-ci * batched : fix n_seq_id * sampling : fix malloc ggml-ci * swift : fix build ggml-ci * swift : try to fix build ggml-ci * prompts : add assistant.txt * common : add llama_batch_add() and llama_batch_clear() helpers * speculative : minor refactor ggml-ci * minor : comments + rename ggml-ci * speculative : fix off-by-one for n_drafted * speculative : fix the n_drafted fix + p constants
2023-10-18 13:21:57 +00:00
llama_batch batch = { 0, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, 0, 0, 0, };
llama : Metal inference (#1642) * mtl : export the LLaMA computation graph * ci : disable temporary * mtl : adapt the MNIST example as starter * mtl : no need for mtl-export tool, add cli arg for main instead * mtl : export just a small part of the graph for now to make it easier * mtl : move MSL code into separate file for easy editing * mtl : initial get_rows_q4_0 kernel * mtl : confirmed get_rows_q4_0 is working correctly * mtl : add rms_norm kernel + confirm working * mtl : add mul kernel + confirm working * mtl : initial mul_mat Q4 kernel (wrong results) * mtl : mul_mat fixes (still wrong) * mtl : another mul_mat Q4 (still does not work) * mtl : working mul_mat q4 * ggml : fix handling of "view" ops in ggml_graph_import() * mtl : add rope kernel * mtl : add reshape and transpose handling * ggml : store offset as opt arg for ggml_view_xd() operators * mtl : add cpy kernel + handle view ops * mtl : confirm f16 x f32 attention mul mat * mtl : add scale kernel * mtl : add diag_mask_inf kernel * mtl : fix soft_max kernel * ggml : update ggml_nbytes() to handle non-contiguous tensors * mtl : verify V tensor contents * mtl : add f32 -> f32 cpy kernel * mtl : add silu kernel * mtl : add non-broadcast mul kernel * mtl : full GPU inference of the computation graph * mtl : optimize rms_norm and soft_max kernels * mtl : add f16 mat x f32 vec multiplication kernel * mtl : fix bug in f16 x f32 mul mat + speed-up computation * mtl : faster mul_mat_q4_0_f32 kernel * mtl : fix kernel signature + roll inner loop * mtl : more threads for rms_norm + better timing * mtl : remove printfs from inner loop * mtl : simplify implementation * mtl : add save/load vocab to ggml file * mtl : plug Metal inference into llama.cpp (very quick-n-dirty) * mtl : make it work with main example Lots of hacks but at least now it generates text * mtl : preparing for merge * mtl : clean-up ggml mtl interface + suport scratch / inplace * mtl : remove temp / debug code * metal : final refactoring and simplification * Revert "ci : disable temporary" This reverts commit 98c267fc77fe811082f672538fc91bcfc9072d63. * metal : add comments * metal : clean-up stuff, fix typos * readme : add Metal instructions * readme : add example for main
2023-06-04 20:34:30 +00:00
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
if (embd) {
llama : fix memory leak in llama_batch_free (#5252) The llama_batch_init allocates memory for a fixed number of tokens. However, the llama_batch_free only frees memory for the number of tokens that were added to the batch. This change-set uses a null terminated array for the batch seq_id, and frees all the elements until the nullptr is reached. This change-set also changes the name of the first parameter from `n_tokens` to `n_tokens_alloc` to more clearly indicate that this value is the number of tokens allocated to the batch, not the number of tokens in the batch.
2024-02-02 07:20:13 +00:00
batch.embd = (float *) malloc(sizeof(float) * n_tokens_alloc * embd);
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
} else {
llama : fix memory leak in llama_batch_free (#5252) The llama_batch_init allocates memory for a fixed number of tokens. However, the llama_batch_free only frees memory for the number of tokens that were added to the batch. This change-set uses a null terminated array for the batch seq_id, and frees all the elements until the nullptr is reached. This change-set also changes the name of the first parameter from `n_tokens` to `n_tokens_alloc` to more clearly indicate that this value is the number of tokens allocated to the batch, not the number of tokens in the batch.
2024-02-02 07:20:13 +00:00
batch.token = (llama_token *) malloc(sizeof(llama_token) * n_tokens_alloc);
llama : Metal inference (#1642) * mtl : export the LLaMA computation graph * ci : disable temporary * mtl : adapt the MNIST example as starter * mtl : no need for mtl-export tool, add cli arg for main instead * mtl : export just a small part of the graph for now to make it easier * mtl : move MSL code into separate file for easy editing * mtl : initial get_rows_q4_0 kernel * mtl : confirmed get_rows_q4_0 is working correctly * mtl : add rms_norm kernel + confirm working * mtl : add mul kernel + confirm working * mtl : initial mul_mat Q4 kernel (wrong results) * mtl : mul_mat fixes (still wrong) * mtl : another mul_mat Q4 (still does not work) * mtl : working mul_mat q4 * ggml : fix handling of "view" ops in ggml_graph_import() * mtl : add rope kernel * mtl : add reshape and transpose handling * ggml : store offset as opt arg for ggml_view_xd() operators * mtl : add cpy kernel + handle view ops * mtl : confirm f16 x f32 attention mul mat * mtl : add scale kernel * mtl : add diag_mask_inf kernel * mtl : fix soft_max kernel * ggml : update ggml_nbytes() to handle non-contiguous tensors * mtl : verify V tensor contents * mtl : add f32 -> f32 cpy kernel * mtl : add silu kernel * mtl : add non-broadcast mul kernel * mtl : full GPU inference of the computation graph * mtl : optimize rms_norm and soft_max kernels * mtl : add f16 mat x f32 vec multiplication kernel * mtl : fix bug in f16 x f32 mul mat + speed-up computation * mtl : faster mul_mat_q4_0_f32 kernel * mtl : fix kernel signature + roll inner loop * mtl : more threads for rms_norm + better timing * mtl : remove printfs from inner loop * mtl : simplify implementation * mtl : add save/load vocab to ggml file * mtl : plug Metal inference into llama.cpp (very quick-n-dirty) * mtl : make it work with main example Lots of hacks but at least now it generates text * mtl : preparing for merge * mtl : clean-up ggml mtl interface + suport scratch / inplace * mtl : remove temp / debug code * metal : final refactoring and simplification * Revert "ci : disable temporary" This reverts commit 98c267fc77fe811082f672538fc91bcfc9072d63. * metal : add comments * metal : clean-up stuff, fix typos * readme : add Metal instructions * readme : add example for main
2023-06-04 20:34:30 +00:00
}
llama : fix memory leak in llama_batch_free (#5252) The llama_batch_init allocates memory for a fixed number of tokens. However, the llama_batch_free only frees memory for the number of tokens that were added to the batch. This change-set uses a null terminated array for the batch seq_id, and frees all the elements until the nullptr is reached. This change-set also changes the name of the first parameter from `n_tokens` to `n_tokens_alloc` to more clearly indicate that this value is the number of tokens allocated to the batch, not the number of tokens in the batch.
2024-02-02 07:20:13 +00:00
batch.pos = (llama_pos *) malloc(sizeof(llama_pos) * n_tokens_alloc);
batch.n_seq_id = (int32_t *) malloc(sizeof(int32_t) * n_tokens_alloc);
batch.seq_id = (llama_seq_id **) malloc(sizeof(llama_seq_id *) * (n_tokens_alloc + 1));
for (int i = 0; i < n_tokens_alloc; ++i) {
speculative : add tree-based sampling example (#3624) * sampling : one sequence per sampling context ggml-ci * speculative : add tree-based sampling support ggml-ci * speculative : reuse the n_parallel CLI param * speculative : refactor sampling * examples : fix build after sampling refactoring ggml-ci * batched : fix n_seq_id * sampling : fix malloc ggml-ci * swift : fix build ggml-ci * swift : try to fix build ggml-ci * prompts : add assistant.txt * common : add llama_batch_add() and llama_batch_clear() helpers * speculative : minor refactor ggml-ci * minor : comments + rename ggml-ci * speculative : fix off-by-one for n_drafted * speculative : fix the n_drafted fix + p constants
2023-10-18 13:21:57 +00:00
batch.seq_id[i] = (llama_seq_id *) malloc(sizeof(llama_seq_id) * n_seq_max);
}
llama : fix memory leak in llama_batch_free (#5252) The llama_batch_init allocates memory for a fixed number of tokens. However, the llama_batch_free only frees memory for the number of tokens that were added to the batch. This change-set uses a null terminated array for the batch seq_id, and frees all the elements until the nullptr is reached. This change-set also changes the name of the first parameter from `n_tokens` to `n_tokens_alloc` to more clearly indicate that this value is the number of tokens allocated to the batch, not the number of tokens in the batch.
2024-02-02 07:20:13 +00:00
batch.seq_id[n_tokens_alloc] = nullptr;
batch.logits = (int8_t *) malloc(sizeof(int8_t) * n_tokens_alloc);
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
return batch;
}
void llama_batch_free(struct llama_batch batch) {
speculative : add tree-based sampling example (#3624) * sampling : one sequence per sampling context ggml-ci * speculative : add tree-based sampling support ggml-ci * speculative : reuse the n_parallel CLI param * speculative : refactor sampling * examples : fix build after sampling refactoring ggml-ci * batched : fix n_seq_id * sampling : fix malloc ggml-ci * swift : fix build ggml-ci * swift : try to fix build ggml-ci * prompts : add assistant.txt * common : add llama_batch_add() and llama_batch_clear() helpers * speculative : minor refactor ggml-ci * minor : comments + rename ggml-ci * speculative : fix off-by-one for n_drafted * speculative : fix the n_drafted fix + p constants
2023-10-18 13:21:57 +00:00
if (batch.token) free(batch.token);
if (batch.embd) free(batch.embd);
if (batch.pos) free(batch.pos);
if (batch.n_seq_id) free(batch.n_seq_id);
if (batch.seq_id) {
llama : fix memory leak in llama_batch_free (#5252) The llama_batch_init allocates memory for a fixed number of tokens. However, the llama_batch_free only frees memory for the number of tokens that were added to the batch. This change-set uses a null terminated array for the batch seq_id, and frees all the elements until the nullptr is reached. This change-set also changes the name of the first parameter from `n_tokens` to `n_tokens_alloc` to more clearly indicate that this value is the number of tokens allocated to the batch, not the number of tokens in the batch.
2024-02-02 07:20:13 +00:00
for (int i = 0; batch.seq_id[i] != nullptr; ++i) {
speculative : add tree-based sampling example (#3624) * sampling : one sequence per sampling context ggml-ci * speculative : add tree-based sampling support ggml-ci * speculative : reuse the n_parallel CLI param * speculative : refactor sampling * examples : fix build after sampling refactoring ggml-ci * batched : fix n_seq_id * sampling : fix malloc ggml-ci * swift : fix build ggml-ci * swift : try to fix build ggml-ci * prompts : add assistant.txt * common : add llama_batch_add() and llama_batch_clear() helpers * speculative : minor refactor ggml-ci * minor : comments + rename ggml-ci * speculative : fix off-by-one for n_drafted * speculative : fix the n_drafted fix + p constants
2023-10-18 13:21:57 +00:00
free(batch.seq_id[i]);
}
free(batch.seq_id);
}
if (batch.logits) free(batch.logits);
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
}
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
int32_t llama_decode(
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
struct llama_context * ctx,
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
struct llama_batch batch) {
const int ret = llama_decode_internal(*ctx, batch);
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
if (ret < 0) {
LLAMA_LOG_ERROR("%s: failed to decode, ret = %d\n", __func__, ret);
}
return ret;
llama : Metal inference (#1642) * mtl : export the LLaMA computation graph * ci : disable temporary * mtl : adapt the MNIST example as starter * mtl : no need for mtl-export tool, add cli arg for main instead * mtl : export just a small part of the graph for now to make it easier * mtl : move MSL code into separate file for easy editing * mtl : initial get_rows_q4_0 kernel * mtl : confirmed get_rows_q4_0 is working correctly * mtl : add rms_norm kernel + confirm working * mtl : add mul kernel + confirm working * mtl : initial mul_mat Q4 kernel (wrong results) * mtl : mul_mat fixes (still wrong) * mtl : another mul_mat Q4 (still does not work) * mtl : working mul_mat q4 * ggml : fix handling of "view" ops in ggml_graph_import() * mtl : add rope kernel * mtl : add reshape and transpose handling * ggml : store offset as opt arg for ggml_view_xd() operators * mtl : add cpy kernel + handle view ops * mtl : confirm f16 x f32 attention mul mat * mtl : add scale kernel * mtl : add diag_mask_inf kernel * mtl : fix soft_max kernel * ggml : update ggml_nbytes() to handle non-contiguous tensors * mtl : verify V tensor contents * mtl : add f32 -> f32 cpy kernel * mtl : add silu kernel * mtl : add non-broadcast mul kernel * mtl : full GPU inference of the computation graph * mtl : optimize rms_norm and soft_max kernels * mtl : add f16 mat x f32 vec multiplication kernel * mtl : fix bug in f16 x f32 mul mat + speed-up computation * mtl : faster mul_mat_q4_0_f32 kernel * mtl : fix kernel signature + roll inner loop * mtl : more threads for rms_norm + better timing * mtl : remove printfs from inner loop * mtl : simplify implementation * mtl : add save/load vocab to ggml file * mtl : plug Metal inference into llama.cpp (very quick-n-dirty) * mtl : make it work with main example Lots of hacks but at least now it generates text * mtl : preparing for merge * mtl : clean-up ggml mtl interface + suport scratch / inplace * mtl : remove temp / debug code * metal : final refactoring and simplification * Revert "ci : disable temporary" This reverts commit 98c267fc77fe811082f672538fc91bcfc9072d63. * metal : add comments * metal : clean-up stuff, fix typos * readme : add Metal instructions * readme : add example for main
2023-06-04 20:34:30 +00:00
}
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
void llama_synchronize(struct llama_context * ctx) {
ggml_backend_sched_synchronize(ctx->sched);
// FIXME: if multiple single tokens are evaluated without a synchronization,
// the stats will be added to the prompt evaluation stats
// this should only happen when using batch size 1 to evaluate a batch
// add the evaluation to the stats
if (ctx->n_queued_tokens == 1) {
ctx->t_eval_us += ggml_time_us() - ctx->t_compute_start_us;
ctx->n_eval++;
} else if (ctx->n_queued_tokens > 1) {
ctx->t_p_eval_us += ggml_time_us() - ctx->t_compute_start_us;
ctx->n_p_eval += ctx->n_queued_tokens;
}
// get a more accurate load time, upon first eval
if (ctx->n_queued_tokens > 0 && !ctx->has_evaluated_once) {
ctx->t_load_us = ggml_time_us() - ctx->t_start_us;
ctx->has_evaluated_once = true;
}
ctx->n_queued_tokens = 0;
ctx->t_compute_start_us = 0;
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
float * llama_get_logits(struct llama_context * ctx) {
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
llama_synchronize(ctx);
return ctx->logits;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
float * llama_get_logits_ith(struct llama_context * ctx, int32_t i) {
llama : support negative ith in llama_get_ API (#6519) * llama_sampling_sample with default args is more naively usable * Batches populated by either llama_batch_get_one or llama_batch_add work with default args * Previously get_one could use the default argument * Previously add should usually have used the last index where logits[idx] == true * This hopefully encourages the use of llama_batch_add * By giving expected results when using default arguments. * Adds "negative indexing" feature to llama_get_logits_ith and llama_get_embeddings_ith * Believed to work with any currently well behaved program * Default arg now works for both cases (previously would give strange results for add case) * Any non-negative number is unaffected and behaves as previously * Negative arguments were previously invalid. * Implemented as a special case of indexing as suggested by @compilade in https://github.com/ggerganov/llama.cpp/pull/6519 * Fixed mismatch type errors * cited in macOS CI tests * Missed in original updates based on PR feedback in https://github.com/ggerganov/llama.cpp/pull/6519
2024-04-08 13:02:30 +00:00
int32_t j = -1;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
llama_synchronize(ctx);
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
try {
if (ctx->logits == nullptr) {
throw std::runtime_error("no logits");
}
llama : support negative ith in llama_get_ API (#6519) * llama_sampling_sample with default args is more naively usable * Batches populated by either llama_batch_get_one or llama_batch_add work with default args * Previously get_one could use the default argument * Previously add should usually have used the last index where logits[idx] == true * This hopefully encourages the use of llama_batch_add * By giving expected results when using default arguments. * Adds "negative indexing" feature to llama_get_logits_ith and llama_get_embeddings_ith * Believed to work with any currently well behaved program * Default arg now works for both cases (previously would give strange results for add case) * Any non-negative number is unaffected and behaves as previously * Negative arguments were previously invalid. * Implemented as a special case of indexing as suggested by @compilade in https://github.com/ggerganov/llama.cpp/pull/6519 * Fixed mismatch type errors * cited in macOS CI tests * Missed in original updates based on PR feedback in https://github.com/ggerganov/llama.cpp/pull/6519
2024-04-08 13:02:30 +00:00
if (i < 0) {
j = ctx->n_outputs + i;
if (j < 0) {
throw std::runtime_error(format("negative index out of range [0, %d)", ctx->n_outputs));
}
} else if ((size_t) i >= ctx->output_ids.size()) {
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
throw std::runtime_error(format("out of range [0, %lu)", ctx->output_ids.size()));
llama : support negative ith in llama_get_ API (#6519) * llama_sampling_sample with default args is more naively usable * Batches populated by either llama_batch_get_one or llama_batch_add work with default args * Previously get_one could use the default argument * Previously add should usually have used the last index where logits[idx] == true * This hopefully encourages the use of llama_batch_add * By giving expected results when using default arguments. * Adds "negative indexing" feature to llama_get_logits_ith and llama_get_embeddings_ith * Believed to work with any currently well behaved program * Default arg now works for both cases (previously would give strange results for add case) * Any non-negative number is unaffected and behaves as previously * Negative arguments were previously invalid. * Implemented as a special case of indexing as suggested by @compilade in https://github.com/ggerganov/llama.cpp/pull/6519 * Fixed mismatch type errors * cited in macOS CI tests * Missed in original updates based on PR feedback in https://github.com/ggerganov/llama.cpp/pull/6519
2024-04-08 13:02:30 +00:00
} else {
j = ctx->output_ids[i];
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
}
if (j < 0) {
throw std::runtime_error(format("batch.logits[%d] != true", i));
}
llama : support negative ith in llama_get_ API (#6519) * llama_sampling_sample with default args is more naively usable * Batches populated by either llama_batch_get_one or llama_batch_add work with default args * Previously get_one could use the default argument * Previously add should usually have used the last index where logits[idx] == true * This hopefully encourages the use of llama_batch_add * By giving expected results when using default arguments. * Adds "negative indexing" feature to llama_get_logits_ith and llama_get_embeddings_ith * Believed to work with any currently well behaved program * Default arg now works for both cases (previously would give strange results for add case) * Any non-negative number is unaffected and behaves as previously * Negative arguments were previously invalid. * Implemented as a special case of indexing as suggested by @compilade in https://github.com/ggerganov/llama.cpp/pull/6519 * Fixed mismatch type errors * cited in macOS CI tests * Missed in original updates based on PR feedback in https://github.com/ggerganov/llama.cpp/pull/6519
2024-04-08 13:02:30 +00:00
if (j >= ctx->n_outputs) {
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
// This should not happen
llama : support negative ith in llama_get_ API (#6519) * llama_sampling_sample with default args is more naively usable * Batches populated by either llama_batch_get_one or llama_batch_add work with default args * Previously get_one could use the default argument * Previously add should usually have used the last index where logits[idx] == true * This hopefully encourages the use of llama_batch_add * By giving expected results when using default arguments. * Adds "negative indexing" feature to llama_get_logits_ith and llama_get_embeddings_ith * Believed to work with any currently well behaved program * Default arg now works for both cases (previously would give strange results for add case) * Any non-negative number is unaffected and behaves as previously * Negative arguments were previously invalid. * Implemented as a special case of indexing as suggested by @compilade in https://github.com/ggerganov/llama.cpp/pull/6519 * Fixed mismatch type errors * cited in macOS CI tests * Missed in original updates based on PR feedback in https://github.com/ggerganov/llama.cpp/pull/6519
2024-04-08 13:02:30 +00:00
throw std::runtime_error(format("corrupt output buffer (j=%d, n_outputs=%d)", j, ctx->n_outputs));
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
}
return ctx->logits + j*ctx->model.hparams.n_vocab;
} catch (const std::exception & err) {
LLAMA_LOG_ERROR("%s: invalid logits id %d, reason: %s\n", __func__, i, err.what());
#ifndef NDEBUG
GGML_ASSERT(false);
#endif
return nullptr;
}
llama : custom attention mask + parallel decoding + no context swaps (#3228) * tests : verify that RoPE is "additive" * llama : replace ggml_diag_mask_inf with ggml_add (custom -inf mask) * ggml : ggml_rope now takes a vector with positions instead of n_past * metal : add rope_f16 kernel + optimize cpy kernels * llama : unified KV cache + batch inference API * llama : add new llama_decode() API that works with llama_batch * llama : add cell_max heuristic for more efficient kv_cache * llama : extend llama_kv_cache API * llama : more robust cell_max heuristic + wip shift * metal : disable concurrency optimization * llama : add llama_kv_cache_shift_seq + no more context swaps * llama : apply K-cache roping for Falcon and Baichuan * speculative : fix KV cache management * parallel : example for serving multiple users in parallel * parallel : disable hot-plug to avoid cache fragmentation * fixes : speculative KV cache + llama worst-case graph * llama : extend batch API to select which logits to output * llama : fix worst case graph build * ggml-cuda : update rope implementation for parallel decoding (#3254) * ggml-cuda : update rope implementation for parallel decoding * better solution for p0 computation * fix rope * simpler rope implementation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * make : add parallel to build + fix static functions in llama.cpp * simple : fix token counting * parallel : various improvements * llama : fix cell_max logic + rename functions * parallel : try smaller batches when the KV cache is fragmented * parallel : fix sequence termination criteria * llama : silence errors KV cache errors * parallel : remove new line from prompt * parallel : process system prompt once + configurable paramters + llama API * parallel : remove question with short answers * parallel : count cache misses * parallel : print misses on each request * parallel : minor * llama : fix n_kv to never become 0 * parallel : rename hot-plug to continuous-batching * llama : improve llama_batch API + simplify parallel example * simple : add parallel decoding support * simple : improve comments + free batch * ggml-cuda : add rope f16, restore performance with parallel decoding (#3272) * ggml-cuda : add rope f16, restore performance * offload KQ_mask with all models * fix rope shift --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : disable MPI for now ggml-ci * train : make KQ_pos memory buffer permanent via dummy scale op * ggml : revert change to ggml_cpy, add ggml_cont_Nd instead (#3275) ggml-ci * parallel : fix bug (extra BOS) + smaller token_prev array * parallel : fix cases where the input prompts can overflow the batch * parallel : add disabled experimental batch chunking in powers of two * llama : llama.h formatting + comments * simple : add README.md * llama : fix kv cache heuristic when context is less than 32 * parallel : fix crash when `-n -1` * llama : simplify returns if/else branches * metal : use mm kernels for batch size > 2 * examples : utilize new llama_get_logits_ith() * examples : add example for batched decoding * examples : do not eval prompt 2 times (close #3348) * server : clear the KV cache beyond n_past before llama_decode * server : avoid context swaps by shifting the KV cache --------- Co-authored-by: slaren <slarengh@gmail.com>
2023-09-28 16:04:36 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
float * llama_get_embeddings(struct llama_context * ctx) {
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
llama_synchronize(ctx);
return ctx->embd;
llama : add functions that work directly on model (#2197) * Remove vocab reference from context * Add functions that works directly with model
2023-07-14 18:55:24 +00:00
}
llama : support batched embeddings (#5466) * batched embedding: pool outputs by sequence id. updated embedding example * bring back non-causal attention * embd : minor improvements * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-13 12:06:58 +00:00
float * llama_get_embeddings_ith(struct llama_context * ctx, int32_t i) {
llama : support negative ith in llama_get_ API (#6519) * llama_sampling_sample with default args is more naively usable * Batches populated by either llama_batch_get_one or llama_batch_add work with default args * Previously get_one could use the default argument * Previously add should usually have used the last index where logits[idx] == true * This hopefully encourages the use of llama_batch_add * By giving expected results when using default arguments. * Adds "negative indexing" feature to llama_get_logits_ith and llama_get_embeddings_ith * Believed to work with any currently well behaved program * Default arg now works for both cases (previously would give strange results for add case) * Any non-negative number is unaffected and behaves as previously * Negative arguments were previously invalid. * Implemented as a special case of indexing as suggested by @compilade in https://github.com/ggerganov/llama.cpp/pull/6519 * Fixed mismatch type errors * cited in macOS CI tests * Missed in original updates based on PR feedback in https://github.com/ggerganov/llama.cpp/pull/6519
2024-04-08 13:02:30 +00:00
int32_t j = -1;
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
llama_synchronize(ctx);
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
try {
if (ctx->embd == nullptr) {
throw std::runtime_error("no embeddings");
}
llama : support negative ith in llama_get_ API (#6519) * llama_sampling_sample with default args is more naively usable * Batches populated by either llama_batch_get_one or llama_batch_add work with default args * Previously get_one could use the default argument * Previously add should usually have used the last index where logits[idx] == true * This hopefully encourages the use of llama_batch_add * By giving expected results when using default arguments. * Adds "negative indexing" feature to llama_get_logits_ith and llama_get_embeddings_ith * Believed to work with any currently well behaved program * Default arg now works for both cases (previously would give strange results for add case) * Any non-negative number is unaffected and behaves as previously * Negative arguments were previously invalid. * Implemented as a special case of indexing as suggested by @compilade in https://github.com/ggerganov/llama.cpp/pull/6519 * Fixed mismatch type errors * cited in macOS CI tests * Missed in original updates based on PR feedback in https://github.com/ggerganov/llama.cpp/pull/6519
2024-04-08 13:02:30 +00:00
if (i < 0) {
j = ctx->n_outputs + i;
if (j < 0) {
throw std::runtime_error(format("negative index out of range [0, %d)", ctx->n_outputs));
}
} else if ((size_t) i >= ctx->output_ids.size()) {
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
throw std::runtime_error(format("out of range [0, %lu)", ctx->output_ids.size()));
llama : support negative ith in llama_get_ API (#6519) * llama_sampling_sample with default args is more naively usable * Batches populated by either llama_batch_get_one or llama_batch_add work with default args * Previously get_one could use the default argument * Previously add should usually have used the last index where logits[idx] == true * This hopefully encourages the use of llama_batch_add * By giving expected results when using default arguments. * Adds "negative indexing" feature to llama_get_logits_ith and llama_get_embeddings_ith * Believed to work with any currently well behaved program * Default arg now works for both cases (previously would give strange results for add case) * Any non-negative number is unaffected and behaves as previously * Negative arguments were previously invalid. * Implemented as a special case of indexing as suggested by @compilade in https://github.com/ggerganov/llama.cpp/pull/6519 * Fixed mismatch type errors * cited in macOS CI tests * Missed in original updates based on PR feedback in https://github.com/ggerganov/llama.cpp/pull/6519
2024-04-08 13:02:30 +00:00
} else {
j = ctx->output_ids[i];
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
}
if (j < 0) {
throw std::runtime_error(format("batch.logits[%d] != true", i));
}
llama : support negative ith in llama_get_ API (#6519) * llama_sampling_sample with default args is more naively usable * Batches populated by either llama_batch_get_one or llama_batch_add work with default args * Previously get_one could use the default argument * Previously add should usually have used the last index where logits[idx] == true * This hopefully encourages the use of llama_batch_add * By giving expected results when using default arguments. * Adds "negative indexing" feature to llama_get_logits_ith and llama_get_embeddings_ith * Believed to work with any currently well behaved program * Default arg now works for both cases (previously would give strange results for add case) * Any non-negative number is unaffected and behaves as previously * Negative arguments were previously invalid. * Implemented as a special case of indexing as suggested by @compilade in https://github.com/ggerganov/llama.cpp/pull/6519 * Fixed mismatch type errors * cited in macOS CI tests * Missed in original updates based on PR feedback in https://github.com/ggerganov/llama.cpp/pull/6519
2024-04-08 13:02:30 +00:00
if (j >= ctx->n_outputs) {
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
// This should not happen
llama : support negative ith in llama_get_ API (#6519) * llama_sampling_sample with default args is more naively usable * Batches populated by either llama_batch_get_one or llama_batch_add work with default args * Previously get_one could use the default argument * Previously add should usually have used the last index where logits[idx] == true * This hopefully encourages the use of llama_batch_add * By giving expected results when using default arguments. * Adds "negative indexing" feature to llama_get_logits_ith and llama_get_embeddings_ith * Believed to work with any currently well behaved program * Default arg now works for both cases (previously would give strange results for add case) * Any non-negative number is unaffected and behaves as previously * Negative arguments were previously invalid. * Implemented as a special case of indexing as suggested by @compilade in https://github.com/ggerganov/llama.cpp/pull/6519 * Fixed mismatch type errors * cited in macOS CI tests * Missed in original updates based on PR feedback in https://github.com/ggerganov/llama.cpp/pull/6519
2024-04-08 13:02:30 +00:00
throw std::runtime_error(format("corrupt output buffer (j=%d, n_outputs=%d)", j, ctx->n_outputs));
llama : greatly reduce output buffer memory usage (#6122) * llama : greatly reduce logits memory usage * llama : more compact state saving and reloading * llama : fix lctx.n_outputs not being set before building graph * perplexity : adapt to the logits API changes * perplexity : fix Winogrande, use correct logits for second choice start The first logits used to evaluate the second choice were not from the end of the common prefix; instead, they were the logits from the end of the first choice. This has been corrected. The previous implementation sometimes had outliers in the scores of choices for some tasks, and the logic to skip choices words in the log-likelihood evaluation probably was an attempt to reduce those, but it was complex and didn't quite seem to be the right thing. This is simpler now, and the outlier scores aren't there anymore. * perplexity : normalize spaces and punctuation in Winogrande sentences * llama : fix embedding conditions * llama : fix llama_get_embeddings_ith when the resulting id is 0 * llama : fix wrong n_outputs in llama_set_inputs A mismatch happened when using a smaller n_ubatch than n_batch and then using llama_batch_get_one(). The decision of what n_outputs should be now almost fully depends on how lctx.n_outputs is set in llama_decode_internal. The conditions are simpler this way. * llama : when saving the state, recalculate n_outputs This ensures the correct number of outputs for the entire previous batch is stored in the session file, even when n_ubatch is smaller than n_batch. * llama : fix not-skipping outputs of non-causal models * llama : fix running a batch with n_outputs == 0 It previously worked because lctx.inp_out_ids was not initialized, so it pointed to some garbage address which was somehow still valid when I ran my tests. * llama : keep same graph topology even when n_outputs == 0 * ggml : saner ggml_can_repeat with empty tensors * ggml : future-proof ggml_is_empty by using GGML_MAX_DIMS - 1 * ggml : do not multi-thread ops returning empty tensors * ggml : make ggml_is_empty public and work with views * llama : use a vector for ctx->output_ids * llama : rework reallocation logic for llama_output_reserve Now comparing the actual size with the new total size of the output buffer to allow more efficient enabling and disabling of the embeddings and/or logits output in the future. * ggml : skip empty tensors in all backends * llama : fix llama_output_reserve nullptr deref when new_size is 0 * perplexity : make Winogrande work as it does on master The problems with the Winogrande implementation will need to be fixed in a separate PR to ease review. * llama : clearer error messages for invalid logits or embeddings ids * llama : assert all models that can have inp_out_ids Since the graph topology is now constant, this presence check can be done even when there are no outputs. * llama : assert logits and embd buffers exist before writing to them * llama : handle errors from llama_output_reserve at call sites * perplexity : make hellaswag and multiple-choice outputs identical to master Due to how the KV cache is updated, the logprobs for tokens in a batch are very slightly affected by the other tokens present in the batch, so to make hellaswag and multiple-choice return exactly the same results as on master, the last token of each sequence needs to be evaluated even though its output is not used at all. This will probably be changed back in the future to make these benchmarks a tiny bit faster. * perplexity : fix division by zero when using less than 100 multiple-choice tasks * llama : allow loading state saved with a different ctx size When loading a session file, the context size is now only required to be at least enough to load the KV cells contained in that session file, instead of requiring to use exactly the same context size as when saving. Doing this enables the use-case of extending or shrinking the context size of a saved session. This breaks existing session files because the meaning of kv_buf_size is slightly changed (previously it was the size of the whole KV cache, now it's only the size of the saved part of it). This allows for finer-grained sanity checks when loading in an effort to keep kv_buf_size useful even when the kv_size is changed. * llama : minor ggml-ci * readme : update recent API changes, and warn about Vulkan --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-26 14:46:41 +00:00
}
return ctx->embd + j*ctx->model.hparams.n_embd;
} catch (const std::exception & err) {
LLAMA_LOG_ERROR("%s: invalid embeddings id %d, reason: %s\n", __func__, i, err.what());
#ifndef NDEBUG
GGML_ASSERT(false);
#endif
return nullptr;
}
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
}
float * llama_get_embeddings_seq(struct llama_context * ctx, llama_seq_id seq_id) {
llama : add pipeline parallelism support (#6017) * llama : add pipeline parallelism support for batch processing with multiple CUDA GPUs ggml-ci * server : add -ub, --ubatch-size parameter * fix server embedding test * llama : fix Mamba inference for pipeline parallelism Tested to work correctly with both `main` and `parallel` examples. * llama : limit max batch size to n_batch * add LLAMA_SCHED_MAX_COPIES to configure the number of input copies for pipeline parallelism default increase to 4 (from 2) changing this value may improve performance for some systems, but increases memory usage * fix hip build * fix sycl build (disable cpy_tensor_async) * fix hip build * llama : limit n_batch and n_ubatch to n_ctx during context creation * llama : fix norm backend * batched-bench : sync after decode * swiftui : sync after decode * ggml : allow ggml_get_rows to use multiple threads if they are available * check n_ubatch >= n_tokens with non-casual attention * llama : do not limit n_batch to n_ctx with non-casual attn * server : construct batch with size of llama_n_batch * ggml_backend_cpu_graph_compute : fix return value when alloc fails * llama : better n_batch and n_ubatch comment * fix merge * small fix * reduce default n_batch to 2048 --------- Co-authored-by: Francis Couture-Harpin <git@compilade.net> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-13 17:54:21 +00:00
llama_synchronize(ctx);
llama : fix embeddings (#5796) * llama : fix embeddings ggml-ci * llama : do not use KV cache for non-causal models ggml-ci * embeddings : fix llama_batch_init arg * llama : add pooling switch * llama : distinguish token vs sequence embeddings ggml-ci * llama : assert pooling tensor * llama : simplify causal mask condition ggml-ci * llama : assert input batch with pooling enabled * readme : update API changes list
2024-03-04 20:31:20 +00:00
auto it = ctx->embd_seq.find(seq_id);
if (it == ctx->embd_seq.end()) {
return nullptr;
}
return it->second.data();
llama : support batched embeddings (#5466) * batched embedding: pool outputs by sequence id. updated embedding example * bring back non-causal attention * embd : minor improvements * llama : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-13 12:06:58 +00:00
}
llama : remove token functions with `context` args in favor of `model` (#3720) * added `llama_model_token_*` variants to all the `llama_token_*` functions. * added `LLAMA_API` * formatting Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * removed old `llama_token` functions * changed 3 more functions to take in model - `llama_token_get_text` - `llama_token_get_score` - `llama_token_get_type` * added back docs * fixed main.cpp * changed token functions to use new model variants * changed token functions to use new model variants --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-23 19:40:03 +00:00
const char * llama_token_get_text(const struct llama_model * model, llama_token token) {
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
GGML_ASSERT(model->vocab.type != LLAMA_VOCAB_TYPE_NONE);
llama : remove token functions with `context` args in favor of `model` (#3720) * added `llama_model_token_*` variants to all the `llama_token_*` functions. * added `LLAMA_API` * formatting Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * removed old `llama_token` functions * changed 3 more functions to take in model - `llama_token_get_text` - `llama_token_get_score` - `llama_token_get_type` * added back docs * fixed main.cpp * changed token functions to use new model variants * changed token functions to use new model variants --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-23 19:40:03 +00:00
return model->vocab.id_to_token[token].text.c_str();
llama : add functions that work directly on model (#2197) * Remove vocab reference from context * Add functions that works directly with model
2023-07-14 18:55:24 +00:00
}
llama : remove token functions with `context` args in favor of `model` (#3720) * added `llama_model_token_*` variants to all the `llama_token_*` functions. * added `LLAMA_API` * formatting Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * removed old `llama_token` functions * changed 3 more functions to take in model - `llama_token_get_text` - `llama_token_get_score` - `llama_token_get_type` * added back docs * fixed main.cpp * changed token functions to use new model variants * changed token functions to use new model variants --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-23 19:40:03 +00:00
float llama_token_get_score(const struct llama_model * model, llama_token token) {
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
GGML_ASSERT(model->vocab.type != LLAMA_VOCAB_TYPE_NONE);
llama : remove token functions with `context` args in favor of `model` (#3720) * added `llama_model_token_*` variants to all the `llama_token_*` functions. * added `LLAMA_API` * formatting Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * removed old `llama_token` functions * changed 3 more functions to take in model - `llama_token_get_text` - `llama_token_get_score` - `llama_token_get_type` * added back docs * fixed main.cpp * changed token functions to use new model variants * changed token functions to use new model variants --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-23 19:40:03 +00:00
return model->vocab.id_to_token[token].score;
llama : add functions that work directly on model (#2197) * Remove vocab reference from context * Add functions that works directly with model
2023-07-14 18:55:24 +00:00
}
llama : remove token functions with `context` args in favor of `model` (#3720) * added `llama_model_token_*` variants to all the `llama_token_*` functions. * added `LLAMA_API` * formatting Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * removed old `llama_token` functions * changed 3 more functions to take in model - `llama_token_get_text` - `llama_token_get_score` - `llama_token_get_type` * added back docs * fixed main.cpp * changed token functions to use new model variants * changed token functions to use new model variants --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-23 19:40:03 +00:00
llama_token_type llama_token_get_type(const struct llama_model * model, llama_token token) {
llama : support models without vocabulary (#5798) * additional methods to read model and ctx parameters * vocab size as a part of a model metadata * models without vocabulary, convert.py part * models without vocabulary, llama.cpp part * PR clean up * converter scrypt fixes * llama_vocab_type update (renamed the new key) * pr review fixes * revert function renaming * one more NoVocab assert
2024-03-14 16:21:56 +00:00
GGML_ASSERT(model->vocab.type != LLAMA_VOCAB_TYPE_NONE);
llama : remove token functions with `context` args in favor of `model` (#3720) * added `llama_model_token_*` variants to all the `llama_token_*` functions. * added `LLAMA_API` * formatting Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * removed old `llama_token` functions * changed 3 more functions to take in model - `llama_token_get_text` - `llama_token_get_score` - `llama_token_get_type` * added back docs * fixed main.cpp * changed token functions to use new model variants * changed token functions to use new model variants --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-23 19:40:03 +00:00
return model->vocab.id_to_token[token].type;
llama : add functions that work directly on model (#2197) * Remove vocab reference from context * Add functions that works directly with model
2023-07-14 18:55:24 +00:00
}
llama : support Llama 3 HF conversion (#6745) * Support Llama 3 conversion The tokenizer is BPE. * style * Accept suggestion Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> * llama : add llama_token_is_eog() ggml-ci * llama : auto-detect more EOT tokens when missing in KV data * convert : replacing EOS token is a hack * llama : fix codegemma EOT token + add TODOs * llama : fix model type string for 8B model --------- Co-authored-by: Sourab Mangrulkar <13534540+pacman100@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-21 11:50:41 +00:00
bool llama_token_is_eog(const struct llama_model * model, llama_token token) {
return token != -1 && (
token == llama_token_eos(model) ||
token == llama_token_eot(model)
);
}
llama : remove token functions with `context` args in favor of `model` (#3720) * added `llama_model_token_*` variants to all the `llama_token_*` functions. * added `LLAMA_API` * formatting Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * removed old `llama_token` functions * changed 3 more functions to take in model - `llama_token_get_text` - `llama_token_get_score` - `llama_token_get_type` * added back docs * fixed main.cpp * changed token functions to use new model variants * changed token functions to use new model variants --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-23 19:40:03 +00:00
llama_token llama_token_bos(const struct llama_model * model) {
return model->vocab.special_bos_id;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
llama : remove token functions with `context` args in favor of `model` (#3720) * added `llama_model_token_*` variants to all the `llama_token_*` functions. * added `LLAMA_API` * formatting Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * removed old `llama_token` functions * changed 3 more functions to take in model - `llama_token_get_text` - `llama_token_get_score` - `llama_token_get_type` * added back docs * fixed main.cpp * changed token functions to use new model variants * changed token functions to use new model variants --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-23 19:40:03 +00:00
llama_token llama_token_eos(const struct llama_model * model) {
return model->vocab.special_eos_id;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
BERT tokenizer fixes (#6498) Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
2024-04-09 17:44:08 +00:00
llama_token llama_token_cls(const struct llama_model * model) {
return model->vocab.special_cls_id;
}
llama_token llama_token_sep(const struct llama_model * model) {
return model->vocab.special_sep_id;
}
llama : remove token functions with `context` args in favor of `model` (#3720) * added `llama_model_token_*` variants to all the `llama_token_*` functions. * added `LLAMA_API` * formatting Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * removed old `llama_token` functions * changed 3 more functions to take in model - `llama_token_get_text` - `llama_token_get_score` - `llama_token_get_type` * added back docs * fixed main.cpp * changed token functions to use new model variants * changed token functions to use new model variants --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-23 19:40:03 +00:00
llama_token llama_token_nl(const struct llama_model * model) {
return model->vocab.linefeed_id;
Cleanup STL headers + fix embedding examples + minor stuff
2023-03-25 18:51:14 +00:00
}
llama : remove token functions with `context` args in favor of `model` (#3720) * added `llama_model_token_*` variants to all the `llama_token_*` functions. * added `LLAMA_API` * formatting Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * removed old `llama_token` functions * changed 3 more functions to take in model - `llama_token_get_text` - `llama_token_get_score` - `llama_token_get_type` * added back docs * fixed main.cpp * changed token functions to use new model variants * changed token functions to use new model variants --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-23 19:40:03 +00:00
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
int32_t llama_add_bos_token(const struct llama_model * model) {
Respect tokenizer.ggml.add_bos_token value when tokenizing (#4040) * gguf-py: gguf-dump: Respect --no-tensor flag in JSON mode. * Respect add_bos_token GGUF metadata value * gguf-py: Try to fix SpecialVocab giving up too easily for the Nth time
2023-11-17 02:14:37 +00:00
return model->vocab.special_add_bos;
}
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
int32_t llama_add_eos_token(const struct llama_model * model) {
Respect tokenizer.ggml.add_bos_token value when tokenizing (#4040) * gguf-py: gguf-dump: Respect --no-tensor flag in JSON mode. * Respect add_bos_token GGUF metadata value * gguf-py: Try to fix SpecialVocab giving up too easily for the Nth time
2023-11-17 02:14:37 +00:00
return model->vocab.special_add_eos;
}
llama : remove token functions with `context` args in favor of `model` (#3720) * added `llama_model_token_*` variants to all the `llama_token_*` functions. * added `LLAMA_API` * formatting Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * removed old `llama_token` functions * changed 3 more functions to take in model - `llama_token_get_text` - `llama_token_get_score` - `llama_token_get_type` * added back docs * fixed main.cpp * changed token functions to use new model variants * changed token functions to use new model variants --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-23 19:40:03 +00:00
llama_token llama_token_prefix(const struct llama_model * model) {
return model->vocab.special_prefix_id;
infill : add new example + extend server API (#3296) * vvhg-code-infill (#1) * infill in separate example (#2) * reverted changes to main and added infill example * cleanup * naming improvement * make : add missing blank line * fix missing semicolon * brought infill up to current main code * cleanup --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com>
2023-10-02 07:42:02 +00:00
}
llama : remove token functions with `context` args in favor of `model` (#3720) * added `llama_model_token_*` variants to all the `llama_token_*` functions. * added `LLAMA_API` * formatting Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * removed old `llama_token` functions * changed 3 more functions to take in model - `llama_token_get_text` - `llama_token_get_score` - `llama_token_get_type` * added back docs * fixed main.cpp * changed token functions to use new model variants * changed token functions to use new model variants --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-23 19:40:03 +00:00
llama_token llama_token_middle(const struct llama_model * model) {
return model->vocab.special_middle_id;
infill : add new example + extend server API (#3296) * vvhg-code-infill (#1) * infill in separate example (#2) * reverted changes to main and added infill example * cleanup * naming improvement * make : add missing blank line * fix missing semicolon * brought infill up to current main code * cleanup --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com>
2023-10-02 07:42:02 +00:00
}
llama : remove token functions with `context` args in favor of `model` (#3720) * added `llama_model_token_*` variants to all the `llama_token_*` functions. * added `LLAMA_API` * formatting Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * removed old `llama_token` functions * changed 3 more functions to take in model - `llama_token_get_text` - `llama_token_get_score` - `llama_token_get_type` * added back docs * fixed main.cpp * changed token functions to use new model variants * changed token functions to use new model variants --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-23 19:40:03 +00:00
llama_token llama_token_suffix(const struct llama_model * model) {
return model->vocab.special_suffix_id;
infill : add new example + extend server API (#3296) * vvhg-code-infill (#1) * infill in separate example (#2) * reverted changes to main and added infill example * cleanup * naming improvement * make : add missing blank line * fix missing semicolon * brought infill up to current main code * cleanup --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com>
2023-10-02 07:42:02 +00:00
}
llama : remove token functions with `context` args in favor of `model` (#3720) * added `llama_model_token_*` variants to all the `llama_token_*` functions. * added `LLAMA_API` * formatting Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * removed old `llama_token` functions * changed 3 more functions to take in model - `llama_token_get_text` - `llama_token_get_score` - `llama_token_get_type` * added back docs * fixed main.cpp * changed token functions to use new model variants * changed token functions to use new model variants --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-23 19:40:03 +00:00
llama_token llama_token_eot(const struct llama_model * model) {
return model->vocab.special_eot_id;
infill : add new example + extend server API (#3296) * vvhg-code-infill (#1) * infill in separate example (#2) * reverted changes to main and added infill example * cleanup * naming improvement * make : add missing blank line * fix missing semicolon * brought infill up to current main code * cleanup --------- Co-authored-by: Cebtenzzre <cebtenzzre@gmail.com>
2023-10-02 07:42:02 +00:00
}
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
int32_t llama_tokenize(
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
const struct llama_model * model,
const char * text,
llama : replace all API facing `int`'s with `int32_t` (#4577) * replaced all API facing `int`'s with `int32_t` * formatting and missed `int` in `llama_token_to_piece`
2024-01-02 14:15:16 +00:00
int32_t text_len,
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama_token * tokens,
llama : more consistent names of count variables (#5994) * llama : more consistent names of count variables ggml-ci * llama : n_parallel -> n_seq_max * common : fix param name * examples : fix param name
2024-03-11 15:49:47 +00:00
int32_t n_tokens_max,
BERT tokenizer fixes (#6498) Key changes: * BERT conversion: fix abuse of LlamaHfVocab, do not set BOS or EOS * Nomic Embed conversion: pad vocab instead of slicing embedding tensor * llama_tokenize: handle added special tokens like HF does
2024-04-09 17:44:08 +00:00
bool add_special,
bool parse_special) {
auto res = llama_tokenize_internal(model->vocab, std::string(text, text_len), add_special, parse_special);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
llama : more consistent names of count variables (#5994) * llama : more consistent names of count variables ggml-ci * llama : n_parallel -> n_seq_max * common : fix param name * examples : fix param name
2024-03-11 15:49:47 +00:00
if (n_tokens_max < (int) res.size()) {
whisper : tokenizer fix + re-enable tokenizer test for LLaMa (#3096) * Fix für #2721 * Reenable tokenizer test for LLaMa * Add `console.cpp` dependency * Fix dependency to `common` * Fixing wrong fix. * Make console usage platform specific Work on compiler warnings. * Adapting makefile * Remove trailing whitespace * Adapting the other parts of the makefile * Fix typo.
2023-09-13 13:19:44 +00:00
// LLAMA_LOG_ERROR("%s: too many tokens\n", __func__);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
return -((int) res.size());
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
for (size_t i = 0; i < res.size(); i++) {
tokens[i] = res[i];
}
llama : add functions that work directly on model (#2197) * Remove vocab reference from context * Add functions that works directly with model
2023-07-14 18:55:24 +00:00
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
return res.size();
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
static std::string llama_decode_text(const std::string & text) {
std::string decoded_text;
tests : add test-tokenizer-0.sh + fix some tokenizers (#7036) * tests : add test-tokenizer-0.sh * unicode : add all unicode number ranges * starcoder : fix pre-tokenizer * tests : add test that fails with DeepSeek tokenizers * falcon : fix regex * unicode : regenerate unicode tables * refact : add tokenizer model * lint : fix * tests : disable failing tests ggml-ci * refact : add tests files ggml-ci * convert : print -> logging ggml-ci * lint : fix * unicode : digit -> number * phi-3 : update
2024-05-04 05:32:32 +00:00
const auto cpts = unicode_cpts_from_utf8(text);
for (const auto cpt : cpts) {
decoded_text += unicode_utf8_to_byte(unicode_cpt_to_utf8(cpt));
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
}
return decoded_text;
}
llama : more tokenizer fixes (#2810) * tests : write a Python tokenizer test (wip) * llama : prefix input text for tokenization with whitespace * llama : distinguish pieces from decoded text + fix detokenization * common : add comments * examples : no longer manually add leading space when tokenizing * tests : use Python to generate tokenizer tests for C++ * tests : add option to tokenize text files ggml-ci * tests : add test-tokenizer-1.py * llama.cpp : fix LF token * hellaswag : move the concat space for clarity * tests : add falcon tests (py + cpp, currently do not pass Unicode) ggml-ci * common : temporary separate llama_detokenize calls for SPM and BPE --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com>
2023-08-27 11:19:19 +00:00
// does not write null-terminator to buf
llama : add option to render special/control tokens (#6807) * make : fix common dep on llama.h * llama : add option to render special tokens * readme : add API change notice ggml-ci * swift : fix build
2024-04-21 15:36:45 +00:00
int32_t llama_token_to_piece(const struct llama_model * model, llama_token token, char * buf, int32_t length, bool special) {
llama.cpp : split llama_context_params into model and context params (#3301) * llama.cpp : split llama_context_params into model and context params ggml-ci * fix metal build * fix freq_base/scale default to model value * llama-bench : keep the same model between tests when possible * move n_threads to llama_context_params, add n_threads_batch * fix mpi build * remove kv_size(), cuda scratch fixes * remove low-vram option * add n_threads_batch to system info, refactor to get_system_info() * add documentation about --threads-batch to the READMEs * llama-bench fix * main : fix rope freq/scale warning * llama.cpp : add llama_get_model common : add llama_tokenize from model * remove duplicated ctx/model functions ggml-ci * cuda : print total VRAM used
2023-09-28 19:42:38 +00:00
if (0 <= token && token < llama_n_vocab(model)) {
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
switch (llama_vocab_get_type(model->vocab)) {
Add support for BERT embedding models (#5423) * BERT model graph construction (build_bert) * WordPiece tokenizer (llm_tokenize_wpm) * Add flag for non-causal attention models * Allow for models that only output embeddings * Support conversion of BERT models to GGUF * Based on prior work by @xyzhang626 and @skeskinen --------- Co-authored-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: Jared Van Bortel <cebtenzzre@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-11 16:21:38 +00:00
case LLAMA_VOCAB_TYPE_WPM:
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
case LLAMA_VOCAB_TYPE_SPM: {
llama : fix detokenization of non-special added-tokens (#4916) Co-authored-by: goerch <jhr.walter@t-online.de>
2024-01-13 16:47:38 +00:00
// NOTE: we accept all unsupported token types,
// suppressing them like CONTROL tokens.
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
if (llama_is_normal_token(model->vocab, token)) {
std::string result = model->vocab.id_to_token[token].text;
llama : escape all U+2581 in a string (#2750)
2023-08-24 09:26:01 +00:00
llama_unescape_whitespace(result);
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
if (length < (int) result.length()) {
cuda : improve cuda pool efficiency using virtual memory (#4606) * cuda : improve cuda pool efficiency using virtual memory * fix mixtral * fix cmake build * check for vmm support, disable for hip ggml-ci * fix hip build * clarify granularity * move all caps to g_device_caps * refactor error checking * add cuda_pool_alloc, refactor most pool allocations ggml-ci * fix hip build * CUBLAS_TF32_TENSOR_OP_MATH is not a macro * more hip crap * llama : fix msvc warnings * ggml : fix msvc warnings * minor * minor * cuda : fallback to CPU on host buffer alloc fail * Update ggml-cuda.cu Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * Update ggml-cuda.cu Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * ensure allocations are always aligned * act_size -> actual_size --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2023-12-24 13:34:22 +00:00
return -(int) result.length();
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
}
memcpy(buf, result.c_str(), result.length());
return result.length();
llama : add option to render special/control tokens (#6807) * make : fix common dep on llama.h * llama : add option to render special tokens * readme : add API change notice ggml-ci * swift : fix build
2024-04-21 15:36:45 +00:00
} else if (
(llama_is_user_defined_token(model->vocab, token)) ||
(llama_is_control_token (model->vocab, token) && special)) {
llama : fix detokenization of non-special added-tokens (#4916) Co-authored-by: goerch <jhr.walter@t-online.de>
2024-01-13 16:47:38 +00:00
std::string result = model->vocab.id_to_token[token].text;
if (length < (int) result.length()) {
ggml, ci : Windows ARM runner and build fixes (#5979) * windows arm ci * fix `error C2078: too many initializers` with ggml_vld1q_u32 macro for MSVC ARM64 * fix `warning C4146: unary minus operator applied to unsigned type, result still unsigned` * fix `error C2065: '__fp16': undeclared identifier`
2024-03-11 09:28:51 +00:00
return -(int) result.length();
llama : fix detokenization of non-special added-tokens (#4916) Co-authored-by: goerch <jhr.walter@t-online.de>
2024-01-13 16:47:38 +00:00
}
memcpy(buf, result.c_str(), result.length());
return result.length();
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
} else if (llama_is_unknown_token(model->vocab, token)) { // NOLINT
if (length < 3) {
return -3;
}
memcpy(buf, "\xe2\x96\x85", 3);
return 3;
} else if (llama_is_byte_token(model->vocab, token)) {
if (length < 1) {
return -1;
}
buf[0] = llama_token_to_byte(model->vocab, token);
return 1;
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
break;
}
case LLAMA_VOCAB_TYPE_BPE: {
llama : fix detokenization of non-special added-tokens (#4916) Co-authored-by: goerch <jhr.walter@t-online.de>
2024-01-13 16:47:38 +00:00
// NOTE: we accept all unsupported token types,
// suppressing them like CONTROL tokens.
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
if (llama_is_normal_token(model->vocab, token)) {
std::string result = model->vocab.id_to_token[token].text;
result = llama_decode_text(result);
if (length < (int) result.length()) {
cuda : improve cuda pool efficiency using virtual memory (#4606) * cuda : improve cuda pool efficiency using virtual memory * fix mixtral * fix cmake build * check for vmm support, disable for hip ggml-ci * fix hip build * clarify granularity * move all caps to g_device_caps * refactor error checking * add cuda_pool_alloc, refactor most pool allocations ggml-ci * fix hip build * CUBLAS_TF32_TENSOR_OP_MATH is not a macro * more hip crap * llama : fix msvc warnings * ggml : fix msvc warnings * minor * minor * cuda : fallback to CPU on host buffer alloc fail * Update ggml-cuda.cu Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * Update ggml-cuda.cu Co-authored-by: Johannes Gäßler <johannesg@5d6.de> * ensure allocations are always aligned * act_size -> actual_size --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
2023-12-24 13:34:22 +00:00
return -(int) result.length();
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
}
memcpy(buf, result.c_str(), result.length());
return result.length();
llama : add option to render special/control tokens (#6807) * make : fix common dep on llama.h * llama : add option to render special tokens * readme : add API change notice ggml-ci * swift : fix build
2024-04-21 15:36:45 +00:00
} else if (
(llama_is_user_defined_token(model->vocab, token)) ||
(llama_is_control_token (model->vocab, token) && special)) {
llama : fix detokenization of non-special added-tokens (#4916) Co-authored-by: goerch <jhr.walter@t-online.de>
2024-01-13 16:47:38 +00:00
std::string result = model->vocab.id_to_token[token].text;
if (length < (int) result.length()) {
ggml, ci : Windows ARM runner and build fixes (#5979) * windows arm ci * fix `error C2078: too many initializers` with ggml_vld1q_u32 macro for MSVC ARM64 * fix `warning C4146: unary minus operator applied to unsigned type, result still unsigned` * fix `error C2065: '__fp16': undeclared identifier`
2024-03-11 09:28:51 +00:00
return -(int) result.length();
llama : fix detokenization of non-special added-tokens (#4916) Co-authored-by: goerch <jhr.walter@t-online.de>
2024-01-13 16:47:38 +00:00
}
memcpy(buf, result.c_str(), result.length());
return result.length();
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
Work on the BPE tokenizer (#3252) * Work on the BPE tokenizer Tokenizer tests work for Falcon-7B * Try to fix build problem * Fix debug assertion failure * Fix MSVC Unicode BOM problem * Cleanup and an improvement * Fix compiler warning * Cleanup * Test doesn't work over the full range of Unicodes * Update .gitignore and Makefile * Another Makefile rule * Testing Aquila * Moving byte decoding back to `token_to_piece` ... ... because everyone is using it. * Guarding some unusable code pathes * Streamlining code and adding some more assertions Important change: I'm classifying added tokens as control tokens now for BPE. * Adding a comment * Adding another assertion * Fixed vocabulary guarding assertions * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fix PR for recent change * Fix PR for recent change * Fix PR for recent change * Fix for compiler warning * Fixes for more compiler warnings * Remove unused code * Fix initialization of static maps * Add scores and token types back, adapt gptneox * Update llama.cpp Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Update unicode.h Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Ported Starcoder and added some assertions * Fix coding style * Apply @jploski 's fix for missing tokens --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-03 07:16:26 +00:00
break;
}
default:
GGML_ASSERT(false);
gguf : new file format with flexible meta data (beta) (#2398) * gguf : first API pass * gguf : read header + meta data * gguf : read tensor info * gguf : initial model loading - not tested * gguf : add gguf_get_tensor_name() * gguf : do not support passing existing ggml_context to gguf_init * gguf : simplify gguf_get_val * gguf : gguf.c is now part of ggml.c * gguf : read / write sample models * gguf : add comments * refactor : reduce code duplication and better API (#2415) * gguf : expose the gguf_type enum through the API for now * gguf : add array support * gguf.py : some code style changes * convert.py : start a new simplified implementation by removing old stuff * convert.py : remove GGML vocab + other obsolete stuff * GGUF : write tensor (#2426) * WIP: Write tensor * GGUF : Support writing tensors in Python * refactor : rm unused import and upd todos * fix : fix errors upd writing example * rm example.gguf * gitignore *.gguf * undo formatting * gguf : add gguf_find_key (#2438) * gguf.cpp : find key example * ggml.h : add gguf_find_key * ggml.c : add gguf_find_key * gguf : fix writing tensors * gguf : do not hardcode tensor names to read * gguf : write sample tensors to read * gguf : add tokenization constants * quick and dirty conversion example * gguf : fix writing gguf arrays * gguf : write tensors one by one and code reuse * gguf : fix writing gguf arrays * gguf : write tensors one by one * gguf : write tensors one by one * gguf : write tokenizer data * gguf : upd gguf conversion script * Update convert-llama-h5-to-gguf.py * gguf : handle already encoded string * ggml.h : get array str and f32 * ggml.c : get arr str and f32 * gguf.py : support any type * Update convert-llama-h5-to-gguf.py * gguf : fix set is not subscriptable * gguf : update convert-llama-h5-to-gguf.py * constants.py : add layer norm eps * gguf.py : add layer norm eps and merges * ggml.h : increase GGML_MAX_NAME to 64 * ggml.c : add gguf_get_arr_n * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Makefile : add gptneox gguf example * Update convert-llama-h5-to-gguf.py * add gptneox gguf example * Update convert-llama-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-gptneox-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * gguf : support custom alignment value * gguf : fix typo in function call * gguf : mmap tensor data example * fix : update convert-llama-h5-to-gguf.py * Update convert-llama-h5-to-gguf.py * convert-gptneox-h5-to-gguf.py : Special tokens * gptneox-main.cpp : special tokens * Update gptneox-main.cpp * constants.py : special tokens * gguf.py : accumulate kv and tensor info data + special tokens * convert-gptneox-h5-to-gguf.py : accumulate kv and ti + special tokens * gguf : gguf counterpart of llama-util.h * gguf-util.h : update note * convert-llama-h5-to-gguf.py : accumulate kv / ti + special tokens * convert-llama-h5-to-gguf.py : special tokens * Delete gptneox-common.cpp * Delete gptneox-common.h * convert-gptneox-h5-to-gguf.py : gpt2bpe tokenizer * gptneox-main.cpp : gpt2 bpe tokenizer * gpt2 bpe tokenizer (handles merges and unicode) * Makefile : remove gptneox-common * gguf.py : bytesarray for gpt2bpe tokenizer * cmpnct_gpt2bpe.hpp : comments * gguf.py : use custom alignment if present * gguf : minor stuff * Update gptneox-main.cpp * map tensor names * convert-gptneox-h5-to-gguf.py : map tensor names * convert-llama-h5-to-gguf.py : map tensor names * gptneox-main.cpp : map tensor names * gguf : start implementing libllama in GGUF (WIP) * gguf : start implementing libllama in GGUF (WIP) * rm binary commited by mistake * upd .gitignore * gguf : calculate n_mult * gguf : inference with 7B model working (WIP) * gguf : rm deprecated function * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : start implementing gguf_file_saver (WIP) * gguf : add gguf_get_kv_type * gguf : add gguf_get_kv_type * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver (WIP) * gguf : write metadata in gguf_file_saver * gguf : rm references to old file formats * gguf : shorter name for member variable * gguf : rm redundant method * gguf : get rid of n_mult, read n_ff from file * Update gguf_tensor_map.py * Update gptneox-main.cpp * gguf : rm references to old file magics * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : start implementing quantization (WIP) * gguf : quantization is working * gguf : roper closing of file * gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : no need to convert tensors twice * convert-llama-h5-to-gguf.py : no need to convert tensors twice * convert-gptneox-h5-to-gguf.py : simplify nbytes * convert-llama-h5-to-gguf.py : simplify nbytes * gptneox-main.cpp : n_layer --> n_block * constants.py : n_layer --> n_block * gguf.py : n_layer --> n_block * convert-gptneox-h5-to-gguf.py : n_layer --> n_block * convert-llama-h5-to-gguf.py : n_layer --> n_block * gptneox-main.cpp : n_layer --> n_block * Update gguf_tensor_map.py * convert-gptneox-h5-to-gguf.py : load model in parts to save memory * convert-llama-h5-to-gguf.py : load model in parts to save memory * convert : write more metadata for LLaMA * convert : rm quantization version * convert-gptneox-h5-to-gguf.py : add file_type key * gptneox-main.cpp : add file_type key * fix conflicts * gguf : add todos and comments * convert-gptneox-h5-to-gguf.py : tensor name map changes * Create gguf_namemap.py : tensor name map changes * Delete gguf_tensor_map.py * gptneox-main.cpp : tensor name map changes * convert-llama-h5-to-gguf.py : fixes * gguf.py : dont add empty strings * simple : minor style changes * gguf : use UNIX line ending * Create convert-llama-7b-pth-to-gguf.py * llama : sync gguf-llama.cpp with latest llama.cpp (#2608) * llama : sync gguf-llama.cpp with latest llama.cpp * minor : indentation + assert * llama : refactor gguf_buffer and gguf_ctx_buffer * llama : minor * gitignore : add gptneox-main * llama : tokenizer fixes (#2549) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * convert : update convert-new.py with tokenizer fixes (#2614) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * llama : sync gguf-llama with llama (#2613) * llama : sync gguf-llama with llama * tests : fix build + warnings (test-tokenizer-1 still fails) * tests : fix wstring_convert * convert : fix layer names * llama : sync gguf-llama.cpp * convert : update HF converter to new tokenizer voodoo magics * llama : update tokenizer style * convert-llama-h5-to-gguf.py : add token types * constants.py : add token types * gguf.py : add token types * convert-llama-7b-pth-to-gguf.py : add token types * gguf-llama.cpp : fix n_head_kv * convert-llama-h5-to-gguf.py : add 70b gqa support * gguf.py : add tensor data layout * convert-llama-h5-to-gguf.py : add tensor data layout * convert-llama-7b-pth-to-gguf.py : add tensor data layout * gptneox-main.cpp : add tensor data layout * convert-llama-h5-to-gguf.py : clarify the reverse permute * llama : refactor model loading code (#2620) * llama : style formatting + remove helper methods * llama : fix quantization using gguf tool * llama : simplify gguf_file_saver * llama : fix method names * llama : simplify write_header() * llama : no need to pass full file loader to the file saver just gguf_ctx * llama : gguf_file_saver write I32 * llama : refactor tensor names (#2622) * gguf: update tensor names searched in quantization * gguf : define tensor names as constants * gguf : initial write API (not tested yet) * gguf : write to file API (not tested) * gguf : initial write API ready + example * gguf : fix header write * gguf : fixes + simplify example + add ggml_nbytes_pad() * gguf : minor * llama : replace gguf_file_saver with new gguf write API * gguf : streaming support when writing files * gguf : remove oboslete write methods * gguf : remove obosolete gguf_get_arr_xxx API * llama : simplify gguf_file_loader * llama : move hparams and vocab from gguf_file_loader to llama_model_loader * llama : merge gguf-util.h in llama.cpp * llama : reorder definitions in .cpp to match .h * llama : minor simplifications * llama : refactor llama_model_loader (WIP) wip : remove ggml_ctx from llama_model_loader wip : merge gguf_file_loader in llama_model_loader * llama : fix shape prints * llama : fix Windows build + fix norm_rms_eps key * llama : throw error on missing KV paris in model meta data * llama : improve printing + log meta data * llama : switch print order of meta data --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> * gguf : deduplicate (#2629) * gguf : better type names * dedup : CPU + Metal is working * ggml : fix warnings about unused results * llama.cpp : fix line feed and compiler warning * llama : fix strncpy warning + note token_to_str does not write null * llama : restore the original load/save session implementation Will migrate this to GGUF in the future * convert-llama-h5-to-gguf.py : support alt ctx param name * ggml : assert when using ggml_mul with non-F32 src1 * examples : dedup simple --------- Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> * gguf.py : merge all files in gguf.py * convert-new.py : pick #2427 for HF 70B support * examples/gguf : no need to keep q option for quantization any more * llama.cpp : print actual model size * llama.cpp : use ggml_elements() * convert-new.py : output gguf (#2635) * convert-new.py : output gguf (WIP) * convert-new.py : add gguf key-value pairs * llama : add hparams.ctx_train + no longer print ftype * convert-new.py : minor fixes * convert-new.py : vocab-only option should work now * llama : fix tokenizer to use llama_char_to_byte * tests : add new ggml-vocab-llama.gguf * convert-new.py : tensor name mapping * convert-new.py : add map for skipping tensor serialization * convert-new.py : convert script now works * gguf.py : pick some of the refactoring from #2644 * convert-new.py : minor fixes * convert.py : update to support GGUF output * Revert "ci : disable CI temporary to not waste energy" This reverts commit 7e82d25f40386540c2c15226300ad998ecd871ea. * convert.py : n_head_kv optional and .gguf file extension * convert.py : better always have n_head_kv and default it to n_head * llama : sync with recent PRs on master * editorconfig : ignore models folder ggml-ci * ci : update ".bin" to ".gguf" extension ggml-ci * llama : fix llama_model_loader memory leak * gptneox : move as a WIP example * llama : fix lambda capture ggml-ci * ggml : fix bug in gguf_set_kv ggml-ci * common.h : .bin --> .gguf * quantize-stats.cpp : .bin --> .gguf * convert.py : fix HF tensor permuting / unpacking ggml-ci * llama.cpp : typo * llama : throw error if gguf fails to init from file ggml-ci * llama : fix tensor name grepping during quantization ggml-ci * gguf.py : write tensors in a single pass (#2644) * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : single pass for writing tensors + refactoring writer * gguf : style fixes in simple conversion script * gguf : refactor gptneox conversion script * gguf : rename h5 to hf (for HuggingFace) * gguf : refactor pth to gguf conversion script * gguf : rm file_type key and method * gguf.py : fix vertical alignment * gguf.py : indentation --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * convert-gptneox-hf-to-gguf.py : fixes * gguf.py : gptneox mapping * convert-llama-hf-to-gguf.py : fixes * convert-llama-7b-pth-to-gguf.py : fixes * ggml.h : reverse GGUF_MAGIC * gguf.py : reverse GGUF_MAGIC * test-tokenizer-0.cpp : fix warning * llama.cpp : print kv general.name * llama.cpp : get special token kv and linefeed token id * llama : print number of tensors per type + print arch + style * tests : update vocab file with new magic * editorconfig : fix whitespaces * llama : re-order functions * llama : remove C++ API + reorganize common source in /common dir * llama : minor API updates * llama : avoid hardcoded special tokens * llama : fix MPI build ggml-ci * llama : introduce enum llama_vocab_type + remove hardcoded string constants * convert-falcon-hf-to-gguf.py : falcon HF --> gguf conversion, not tested * falcon-main.cpp : falcon inference example * convert-falcon-hf-to-gguf.py : remove extra kv * convert-gptneox-hf-to-gguf.py : remove extra kv * convert-llama-7b-pth-to-gguf.py : remove extra kv * convert-llama-hf-to-gguf.py : remove extra kv * gguf.py : fix for falcon 40b * falcon-main.cpp : fix for falcon 40b * convert-falcon-hf-to-gguf.py : update ref * convert-falcon-hf-to-gguf.py : add tensor data layout * cmpnct_gpt2bpe.hpp : fixes * falcon-main.cpp : fixes * gptneox-main.cpp : fixes * cmpnct_gpt2bpe.hpp : remove non-general stuff * Update examples/server/README.md Co-authored-by: slaren <slarengh@gmail.com> * cmpnct_gpt2bpe.hpp : cleanup * convert-llama-hf-to-gguf.py : special tokens * convert-llama-7b-pth-to-gguf.py : special tokens * convert-permute-debug.py : permute debug print * convert-permute-debug-master.py : permute debug for master * convert-permute-debug.py : change permute type of attn_q * convert.py : 70b model working (change attn_q permute) * Delete convert-permute-debug-master.py * Delete convert-permute-debug.py * convert-llama-hf-to-gguf.py : fix attn_q permute * gguf.py : fix rope scale kv * convert-llama-hf-to-gguf.py : rope scale and added tokens * convert-llama-7b-pth-to-gguf.py : rope scale and added tokens * llama.cpp : use rope scale kv * convert-llama-7b-pth-to-gguf.py : rope scale fix * convert-llama-hf-to-gguf.py : rope scale fix * py : fix whitespace * gguf : add Python script to convert GGMLv3 LLaMA models to GGUF (#2682) * First pass at converting GGMLv3 LLaMA models to GGUF * Cleanups, better output during conversion * Fix vocab space conversion logic * More vocab conversion fixes * Add description to converted GGUF files * Improve help text, expand warning * Allow specifying name and description for output GGUF * Allow overriding vocab and hyperparams from original model metadata * Use correct params override var name * Fix wrong type size for Q8_K Better handling of original style metadata * Set default value for gguf add_tensor raw_shape KW arg * llama : improve token type support (#2668) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * llama : add API for token type ggml-ci * tests : use new tokenizer type API (#2692) * Merge tokenizer fixes into the gguf branch. * Add test vocabularies * Adapt convert-new.py (and fix a clang-cl compiler error on windows) * Improved tokenizer test But does it work on MacOS? * Improve token type support - Added @klosax code to convert.py - Improved token type support in vocabulary * Exclude platform dependent tests * More sentencepiece compatibility by eliminating magic numbers * Restored accidentally removed comment * Improve commentary * Use token type API in test-tokenizer-1.cpp * py : cosmetics * readme : add notice about new file format ggml-ci --------- Co-authored-by: M. Yusuf Sarıgöz <yusufsarigoz@gmail.com> Co-authored-by: klosax <131523366+klosax@users.noreply.github.com> Co-authored-by: goerch <jhr.walter@t-online.de> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Kerfuffle <44031344+KerfuffleV2@users.noreply.github.com>
2023-08-21 20:07:43 +00:00
}
}
return 0;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
llama : add llama_chat_apply_template() (#5538) * llama: add llama_chat_apply_template * test-chat-template: remove dedundant vector * chat_template: do not use std::string for buffer * add clarification for llama_chat_apply_template * llama_chat_apply_template: add zephyr template * llama_chat_apply_template: correct docs * llama_chat_apply_template: use term "chat" everywhere * llama_chat_apply_template: change variable name to "tmpl"
2024-02-19 08:23:37 +00:00
// trim whitespace from the beginning and end of a string
static std::string trim(const std::string & str) {
size_t start = 0;
size_t end = str.size();
while (start < end && isspace(str[start])) {
start += 1;
}
while (end > start && isspace(str[end - 1])) {
end -= 1;
}
return str.substr(start, end - start);
}
// Simple version of "llama_apply_chat_template" that only works with strings
// This function uses heuristic checks to determine commonly used template. It is not a jinja parser.
static int32_t llama_chat_apply_template_internal(
const std::string & tmpl,
minor : fix trailing whitespace (#5538)
2024-02-19 08:34:10 +00:00
const std::vector<const llama_chat_message *> & chat,
llama : add llama_chat_apply_template() (#5538) * llama: add llama_chat_apply_template * test-chat-template: remove dedundant vector * chat_template: do not use std::string for buffer * add clarification for llama_chat_apply_template * llama_chat_apply_template: add zephyr template * llama_chat_apply_template: correct docs * llama_chat_apply_template: use term "chat" everywhere * llama_chat_apply_template: change variable name to "tmpl"
2024-02-19 08:23:37 +00:00
std::string & dest, bool add_ass) {
// Taken from the research: https://github.com/ggerganov/llama.cpp/issues/5527
std::stringstream ss;
add alias for chat template (#5858)
2024-03-04 11:22:08 +00:00
if (tmpl == "chatml" || tmpl.find("<|im_start|>") != std::string::npos) {
llama : add llama_chat_apply_template() (#5538) * llama: add llama_chat_apply_template * test-chat-template: remove dedundant vector * chat_template: do not use std::string for buffer * add clarification for llama_chat_apply_template * llama_chat_apply_template: add zephyr template * llama_chat_apply_template: correct docs * llama_chat_apply_template: use term "chat" everywhere * llama_chat_apply_template: change variable name to "tmpl"
2024-02-19 08:23:37 +00:00
// chatml template
for (auto message : chat) {
ss << "<|im_start|>" << message->role << "\n" << message->content << "<|im_end|>\n";
}
if (add_ass) {
ss << "<|im_start|>assistant\n";
}
add alias for chat template (#5858)
2024-03-04 11:22:08 +00:00
} else if (tmpl == "llama2" || tmpl.find("[INST]") != std::string::npos) {
llama : add llama_chat_apply_template() (#5538) * llama: add llama_chat_apply_template * test-chat-template: remove dedundant vector * chat_template: do not use std::string for buffer * add clarification for llama_chat_apply_template * llama_chat_apply_template: add zephyr template * llama_chat_apply_template: correct docs * llama_chat_apply_template: use term "chat" everywhere * llama_chat_apply_template: change variable name to "tmpl"
2024-02-19 08:23:37 +00:00
// llama2 template and its variants
// [variant] support system message
bool support_system_message = tmpl.find("<<SYS>>") != std::string::npos;
// [variant] space before + after response
bool space_around_response = tmpl.find("' ' + eos_token") != std::string::npos;
// [variant] add BOS inside history
bool add_bos_inside_history = tmpl.find("bos_token + '[INST]") != std::string::npos;
// [variant] trim spaces from the input message
bool strip_message = tmpl.find("content.strip()") != std::string::npos;
// construct the prompt
bool is_inside_turn = true; // skip BOS at the beginning
ss << "[INST] ";
for (auto message : chat) {
std::string content = strip_message ? trim(message->content) : message->content;
std::string role(message->role);
if (!is_inside_turn) {
is_inside_turn = true;
ss << (add_bos_inside_history ? "<s>[INST] " : "[INST] ");
}
if (role == "system") {
if (support_system_message) {
ss << "<<SYS>>\n" << content << "\n<</SYS>>\n\n";
} else {
// if the model does not support system message, we still include it in the first message, but without <<SYS>>
ss << content << "\n";
}
} else if (role == "user") {
ss << content << " [/INST]";
} else {
ss << (space_around_response ? " " : "") << content << (space_around_response ? " " : "") << "</s>";
is_inside_turn = false;
}
}
// llama2 templates seem to not care about "add_generation_prompt"
add alias for chat template (#5858)
2024-03-04 11:22:08 +00:00
} else if (tmpl == "zephyr" || tmpl.find("<|user|>") != std::string::npos) {
llama : add llama_chat_apply_template() (#5538) * llama: add llama_chat_apply_template * test-chat-template: remove dedundant vector * chat_template: do not use std::string for buffer * add clarification for llama_chat_apply_template * llama_chat_apply_template: add zephyr template * llama_chat_apply_template: correct docs * llama_chat_apply_template: use term "chat" everywhere * llama_chat_apply_template: change variable name to "tmpl"
2024-02-19 08:23:37 +00:00
// zephyr template
for (auto message : chat) {
ss << "<|" << message->role << "|>" << "\n" << message->content << "<|endoftext|>\n";
}
if (add_ass) {
ss << "<|assistant|>\n";
}
add alias for chat template (#5858)
2024-03-04 11:22:08 +00:00
} else if (tmpl == "monarch" || tmpl.find("bos_token + message['role']") != std::string::npos) {
server : fallback to chatml, add AlphaMonarch chat template (#5628) * server: fallback to chatml * add new chat template * server: add AlphaMonarch to test chat template * server: only check model template if there is no custom tmpl * remove TODO
2024-02-22 08:33:24 +00:00
// mlabonne/AlphaMonarch-7B template (the <s> is included inside history)
for (auto message : chat) {
std::string bos = (message == chat.front()) ? "" : "<s>"; // skip BOS for first message
ss << bos << message->role << "\n" << message->content << "</s>\n";
}
if (add_ass) {
ss << "<s>assistant\n";
}
add alias for chat template (#5858)
2024-03-04 11:22:08 +00:00
} else if (tmpl == "gemma" || tmpl.find("<start_of_turn>") != std::string::npos) {
Add Gemma chat template (#5665) * add gemma chat template * gemma: only apply system_prompt on non-model message
2024-02-22 18:10:21 +00:00
// google/gemma-7b-it
std::string system_prompt = "";
for (auto message : chat) {
std::string role(message->role);
if (role == "system") {
// there is no system message for gemma, but we will merge it with user prompt, so nothing is broken
system_prompt = trim(message->content);
continue;
}
// in gemma, "assistant" is "model"
role = role == "assistant" ? "model" : message->role;
ss << "<start_of_turn>" << role << "\n";
if (!system_prompt.empty() && role != "model") {
ss << system_prompt << "\n\n";
system_prompt = "";
}
ss << trim(message->content) << "<end_of_turn>\n";
}
if (add_ass) {
ss << "<start_of_turn>model\n";
}
llama : add Orion chat template (#6066)
2024-03-15 08:44:57 +00:00
} else if (tmpl == "orion" || tmpl.find("'\\n\\nAssistant: ' + eos_token") != std::string::npos) {
// OrionStarAI/Orion-14B-Chat
std::string system_prompt = "";
for (auto message : chat) {
std::string role(message->role);
if (role == "system") {
// there is no system message support, we will merge it with user prompt
system_prompt = message->content;
continue;
} else if (role == "user") {
ss << "Human: ";
if (!system_prompt.empty()) {
ss << system_prompt << "\n\n";
system_prompt = "";
}
ss << message->content << "\n\nAssistant: </s>";
} else {
ss << message->content << "</s>";
}
}
Add OpenChat, Alpaca, Vicuna chat templates (#6397) * Add openchat chat template * Add chat template test for openchat * Add chat template for vicuna * Add chat template for orca-vicuna * Add EOS for vicuna templates * Combine vicuna chat templates * Add tests for openchat and vicuna chat templates * Add chat template for alpaca * Add separate template name for vicuna-orca * Remove alpaca, match deepseek with jinja output * Regenerate chat template test with add_generation_prompt * Separate deepseek bos from system message * Match openchat template with jinja output * Remove BOS token from templates, unprefix openchat
2024-04-03 15:24:31 +00:00
} else if (tmpl == "openchat" || tmpl.find("GPT4 Correct ") != std::string::npos) {
// openchat/openchat-3.5-0106,
for (auto message : chat) {
std::string role(message->role);
if (role == "system") {
ss << message->content << "<|end_of_turn|>";
} else {
role[0] = toupper(role[0]);
ss << "GPT4 Correct " << role << ": " << message->content << "<|end_of_turn|>";
}
}
if (add_ass) {
ss << "GPT4 Correct Assistant:";
}
} else if (tmpl == "vicuna" || tmpl == "vicuna-orca" || (tmpl.find("USER: ") != std::string::npos && tmpl.find("ASSISTANT: ") != std::string::npos)) {
// eachadea/vicuna-13b-1.1 (and Orca variant)
for (auto message : chat) {
std::string role(message->role);
if (role == "system") {
// Orca-Vicuna variant uses a system prefix
if (tmpl == "vicuna-orca" || tmpl.find("SYSTEM: ") != std::string::npos) {
ss << "SYSTEM: " << message->content << "\n";
} else {
ss << message->content << "\n\n";
}
} else if (role == "user") {
ss << "USER: " << message->content << "\n";
} else if (role == "assistant") {
ss << "ASSISTANT: " << message->content << "</s>\n";
}
}
if (add_ass) {
ss << "ASSISTANT:";
}
} else if (tmpl == "deepseek" || (tmpl.find("### Instruction:") != std::string::npos && tmpl.find("<|EOT|>") != std::string::npos)) {
// deepseek-ai/deepseek-coder-33b-instruct
for (auto message : chat) {
std::string role(message->role);
if (role == "system") {
ss << message->content;
} else if (role == "user") {
ss << "### Instruction:\n" << message->content << "\n";
} else if (role == "assistant") {
ss << "### Response:\n" << message->content << "\n<|EOT|>\n";
}
}
if (add_ass) {
ss << "### Response:\n";
}
Add Command R chat template (#6650) * Add chat template for command-r model series * Fix indentation * Add chat template test for command-r models and update the implementation to trim whitespaces * Remove debug print
2024-04-14 16:16:34 +00:00
} else if (tmpl == "command-r" || (tmpl.find("<|START_OF_TURN_TOKEN|>") != std::string::npos && tmpl.find("<|USER_TOKEN|>") != std::string::npos)) {
// CohereForAI/c4ai-command-r-plus
for (auto message : chat) {
std::string role(message->role);
if (role == "system") {
ss << "<|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>" << trim(message->content) << "<|END_OF_TURN_TOKEN|>";
} else if (role == "user") {
ss << "<|START_OF_TURN_TOKEN|><|USER_TOKEN|>" << trim(message->content) << "<|END_OF_TURN_TOKEN|>";
} else if (role == "assistant") {
ss << "<|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>" << trim(message->content) << "<|END_OF_TURN_TOKEN|>";
}
}
if (add_ass) {
ss << "<|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>";
}
llama : add llama-3 chat template (#6751) * Added llama-3 chat template * Update llama.cpp Co-authored-by: Samuel Tallet <36248671+SamuelTallet@users.noreply.github.com> * Update llama.cpp Co-authored-by: Samuel Tallet <36248671+SamuelTallet@users.noreply.github.com> * Update tests/test-chat-template.cpp Co-authored-by: Samuel Tallet <36248671+SamuelTallet@users.noreply.github.com> * Added EOS stop sequence according to https://github.com/ggerganov/llama.cpp/pull/6751#issuecomment-2065602862 * Removed adding of BOS token before first message * Removed bos token from expected output from llama-3 * Update tests/test-chat-template.cpp Co-authored-by: Rene Leonhardt <65483435+reneleonhardt@users.noreply.github.com> * Update tests/test-chat-template.cpp Co-authored-by: Rene Leonhardt <65483435+reneleonhardt@users.noreply.github.com> * Added <|end_of_text|> as another stop token * Reverted last change of adding the end_of_text stop word for llama 3 --------- Co-authored-by: Wouter Tichelaar <tichelaarw@spar.net> Co-authored-by: Samuel Tallet <36248671+SamuelTallet@users.noreply.github.com> Co-authored-by: Rene Leonhardt <65483435+reneleonhardt@users.noreply.github.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-21 13:03:39 +00:00
} else if (tmpl == "llama3" || (tmpl.find("<|start_header_id|>") != std::string::npos && tmpl.find("<|end_header_id|>") != std::string::npos)) {
// Llama 3
for (auto message : chat) {
std::string role(message->role);
ss << "<|start_header_id|>" << role << "<|end_header_id|>\n\n" << trim(message->content) << "<|eot_id|>";
}
if (add_ass) {
ss << "<|start_header_id|>assistant<|end_header_id|>\n\n";
}
llama : add phi 3 chat template (#6857) * Add phi 3 chat template & tests * test : fix chat template result --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-04-24 08:52:37 +00:00
} else if (tmpl == "phi3" || (tmpl.find("<|assistant|>") != std::string::npos && tmpl.find("<|end|>") != std::string::npos )) {
// Phi 3
for (auto message : chat) {
std::string role(message->role);
ss << "<|" << role << "|>\n" << trim(message->content) << "<|end|>\n";
}
if (add_ass) {
ss << "<|assistant|>\n";
}
llama : add llama_chat_apply_template() (#5538) * llama: add llama_chat_apply_template * test-chat-template: remove dedundant vector * chat_template: do not use std::string for buffer * add clarification for llama_chat_apply_template * llama_chat_apply_template: add zephyr template * llama_chat_apply_template: correct docs * llama_chat_apply_template: use term "chat" everywhere * llama_chat_apply_template: change variable name to "tmpl"
2024-02-19 08:23:37 +00:00
} else {
// template not supported
return -1;
}
dest = ss.str();
return dest.size();
}
LLAMA_API int32_t llama_chat_apply_template(
const struct llama_model * model,
const char * tmpl,
const struct llama_chat_message * chat,
size_t n_msg,
bool add_ass,
char * buf,
int32_t length) {
std::string curr_tmpl(tmpl == nullptr ? "" : tmpl);
if (tmpl == nullptr) {
GGML_ASSERT(model != nullptr);
// load template from model
std::vector<char> model_template(2048, 0); // longest known template is about 1200 bytes
std::string template_key = "tokenizer.chat_template";
Server: use llama_chat_apply_template (#5593) * server: use llama_chat_apply_template * server: remove trailing space * server: fix format_chat * server: fix help message Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * server: fix formatted_chat --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-02-20 14:58:27 +00:00
int32_t res = llama_model_meta_val_str(model, template_key.c_str(), model_template.data(), model_template.size());
llama : add llama_chat_apply_template() (#5538) * llama: add llama_chat_apply_template * test-chat-template: remove dedundant vector * chat_template: do not use std::string for buffer * add clarification for llama_chat_apply_template * llama_chat_apply_template: add zephyr template * llama_chat_apply_template: correct docs * llama_chat_apply_template: use term "chat" everywhere * llama_chat_apply_template: change variable name to "tmpl"
2024-02-19 08:23:37 +00:00
if (res < 0) {
// worst case: there is no information about template, we will use chatml by default
add alias for chat template (#5858)
2024-03-04 11:22:08 +00:00
curr_tmpl = "chatml"; // see llama_chat_apply_template_internal
llama : add llama_chat_apply_template() (#5538) * llama: add llama_chat_apply_template * test-chat-template: remove dedundant vector * chat_template: do not use std::string for buffer * add clarification for llama_chat_apply_template * llama_chat_apply_template: add zephyr template * llama_chat_apply_template: correct docs * llama_chat_apply_template: use term "chat" everywhere * llama_chat_apply_template: change variable name to "tmpl"
2024-02-19 08:23:37 +00:00
} else {
curr_tmpl = std::string(model_template.data(), model_template.size());
}
}
server : refactor (#5882) * server : refactoring (wip) * server : remove llava/clip objects from build * server : fix empty prompt handling + all slots idle logic * server : normalize id vars * server : code style * server : simplify model chat template validation * server : code style * server : minor * llama : llama_chat_apply_template support null buf * server : do not process embedding requests when disabled * server : reorganize structs and enums + naming fixes * server : merge oai.hpp in utils.hpp * server : refactor system prompt update at start * server : disable cached prompts with self-extend * server : do not process more than n_batch tokens per iter * server: tests: embeddings use a real embeddings model (#5908) * server, tests : bump batch to fit 1 embedding prompt * server: tests: embeddings fix build type Debug is randomly failing (#5911) * server: tests: embeddings, use different KV Cache size * server: tests: embeddings, fixed prompt do not exceed n_batch, increase embedding timeout, reduce number of concurrent embeddings * server: tests: embeddings, no need to wait for server idle as it can timout * server: refactor: clean up http code (#5912) * server : avoid n_available var ggml-ci * server: refactor: better http codes * server : simplify json parsing + add comment about t_last * server : rename server structs * server : allow to override FQDN in tests ggml-ci * server : add comments --------- Co-authored-by: Pierrick Hymbert <pierrick.hymbert@gmail.com>
2024-03-07 09:41:53 +00:00
llama : add llama_chat_apply_template() (#5538) * llama: add llama_chat_apply_template * test-chat-template: remove dedundant vector * chat_template: do not use std::string for buffer * add clarification for llama_chat_apply_template * llama_chat_apply_template: add zephyr template * llama_chat_apply_template: correct docs * llama_chat_apply_template: use term "chat" everywhere * llama_chat_apply_template: change variable name to "tmpl"
2024-02-19 08:23:37 +00:00
// format the chat to string
std::vector<const llama_chat_message *> chat_vec;
chat_vec.resize(n_msg);
for (size_t i = 0; i < n_msg; i++) {
chat_vec[i] = &chat[i];
}
server : refactor (#5882) * server : refactoring (wip) * server : remove llava/clip objects from build * server : fix empty prompt handling + all slots idle logic * server : normalize id vars * server : code style * server : simplify model chat template validation * server : code style * server : minor * llama : llama_chat_apply_template support null buf * server : do not process embedding requests when disabled * server : reorganize structs and enums + naming fixes * server : merge oai.hpp in utils.hpp * server : refactor system prompt update at start * server : disable cached prompts with self-extend * server : do not process more than n_batch tokens per iter * server: tests: embeddings use a real embeddings model (#5908) * server, tests : bump batch to fit 1 embedding prompt * server: tests: embeddings fix build type Debug is randomly failing (#5911) * server: tests: embeddings, use different KV Cache size * server: tests: embeddings, fixed prompt do not exceed n_batch, increase embedding timeout, reduce number of concurrent embeddings * server: tests: embeddings, no need to wait for server idle as it can timout * server: refactor: clean up http code (#5912) * server : avoid n_available var ggml-ci * server: refactor: better http codes * server : simplify json parsing + add comment about t_last * server : rename server structs * server : allow to override FQDN in tests ggml-ci * server : add comments --------- Co-authored-by: Pierrick Hymbert <pierrick.hymbert@gmail.com>
2024-03-07 09:41:53 +00:00
llama : add llama_chat_apply_template() (#5538) * llama: add llama_chat_apply_template * test-chat-template: remove dedundant vector * chat_template: do not use std::string for buffer * add clarification for llama_chat_apply_template * llama_chat_apply_template: add zephyr template * llama_chat_apply_template: correct docs * llama_chat_apply_template: use term "chat" everywhere * llama_chat_apply_template: change variable name to "tmpl"
2024-02-19 08:23:37 +00:00
std::string formatted_chat;
int32_t res = llama_chat_apply_template_internal(curr_tmpl, chat_vec, formatted_chat, add_ass);
if (res < 0) {
return res;
}
server : refactor (#5882) * server : refactoring (wip) * server : remove llava/clip objects from build * server : fix empty prompt handling + all slots idle logic * server : normalize id vars * server : code style * server : simplify model chat template validation * server : code style * server : minor * llama : llama_chat_apply_template support null buf * server : do not process embedding requests when disabled * server : reorganize structs and enums + naming fixes * server : merge oai.hpp in utils.hpp * server : refactor system prompt update at start * server : disable cached prompts with self-extend * server : do not process more than n_batch tokens per iter * server: tests: embeddings use a real embeddings model (#5908) * server, tests : bump batch to fit 1 embedding prompt * server: tests: embeddings fix build type Debug is randomly failing (#5911) * server: tests: embeddings, use different KV Cache size * server: tests: embeddings, fixed prompt do not exceed n_batch, increase embedding timeout, reduce number of concurrent embeddings * server: tests: embeddings, no need to wait for server idle as it can timout * server: refactor: clean up http code (#5912) * server : avoid n_available var ggml-ci * server: refactor: better http codes * server : simplify json parsing + add comment about t_last * server : rename server structs * server : allow to override FQDN in tests ggml-ci * server : add comments --------- Co-authored-by: Pierrick Hymbert <pierrick.hymbert@gmail.com>
2024-03-07 09:41:53 +00:00
if (buf && length > 0) {
strncpy(buf, formatted_chat.c_str(), length);
}
llama : add llama_chat_apply_template() (#5538) * llama: add llama_chat_apply_template * test-chat-template: remove dedundant vector * chat_template: do not use std::string for buffer * add clarification for llama_chat_apply_template * llama_chat_apply_template: add zephyr template * llama_chat_apply_template: correct docs * llama_chat_apply_template: use term "chat" everywhere * llama_chat_apply_template: change variable name to "tmpl"
2024-02-19 08:23:37 +00:00
return res;
}
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
LLAMA_API int llama_split_path(char * split_path, size_t maxlen, const char * path_prefix, int split_no, int split_count) {
static const char * const SPLIT_PATH_FORMAT = "%s-%05d-of-%05d.gguf";
if (snprintf(split_path, maxlen, SPLIT_PATH_FORMAT, path_prefix, split_no + 1, split_count)) {
return strlen(split_path);
}
return 0;
}
int llama_split_prefix(char * dest, size_t maxlen, const char * split_path, int split_no, int split_count) {
std::string str_split_path(split_path);
char postfix[32];
snprintf(postfix, 32, "-%05d-of-%05d.gguf", split_no + 1, split_count);
std::string str_postfix(postfix);
// check if dest ends with postfix
int size_prefix = str_split_path.size() - str_postfix.size();
if (size_prefix > 0 && str_split_path.find(str_postfix, size_prefix) != std::string::npos) {
common: llama_load_model_from_url split support (#6192) * llama: llama_split_prefix fix strncpy does not include string termination common: llama_load_model_from_url: - fix header name case sensitive - support downloading additional split in parallel - hide password in url * common: EOL EOF * common: remove redundant LLAMA_CURL_MAX_PATH_LENGTH definition * common: change max url max length * common: minor comment * server: support HF URL options * llama: llama_model_loader fix log * common: use a constant for max url length * common: clean up curl if file cannot be loaded in gguf * server: tests: add split tests, and HF options params * common: move llama_download_hide_password_in_url inside llama_download_file as a lambda * server: tests: enable back Release test on PR * spacing Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * spacing Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * spacing Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-03-23 17:07:00 +00:00
snprintf(dest, std::min((size_t) size_prefix + 1, maxlen), "%s", split_path);
llama_model_loader: support multiple split/shard GGUFs (#6187) * split: support in llama_model_loader * avoid copying the entire vector Co-authored-by: slaren <slarengh@gmail.com> * split: move llama_tensor_offset to llama_model_loader * llama_model_loader: PR feedbacks: - use only one gguf_context for metadata only - store all ggml_context in a vector as the files and mappings - store all weights in a vector along with the source tensor - rename ctx_gguf to meta - rename ctx_meta to contexts * avoid copying the entire vector * Simplify this by making these optional, switch some layer creation tensor optional Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * Handle optional tensors Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama_model_loader: fail if backend cannot allocate buffer * fix mmap buffer management * llama_model_loader: map file to backend buffer if the allocation succeeds only * llama_model_loader: only map tensors included in the context * llama_model_loader: minor, use same variable name for consistency, fix spacing in types cast * llama_model_loader: fail if any of backend buffer cannot be allocated * spacing Co-authored-by: slaren <slarengh@gmail.com> * fix loop over pointer Co-authored-by: slaren <slarengh@gmail.com> * llama_model_loader: if n_tensors declared not equals to loaded tensors in split, throw an exception instead of asserting * llama_model_loader: ensure mappings vector has the expected size * llama_model_loader: use at instead of operator[] if this should never add to the map. * llama_model_loader: immediately add the backend buffer to the model buffers in order to free them if an error occurs in the next allocation. Reserve the expected size. * llama_model_loader: be sure the model mappings has enough capacity before allocating backend buffer * llama_model_loader: fix map -> unordered map * llama_split_prefix: use a clearer version, not pass split path len but dest max len. Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com> * llama : minor ggml-ci * llama : introduce some typedef helpers * docs: add model shard in hot topic * llama_model_loader: put mapping in a unique_ptr from the moment it is allocated Co-authored-by: slaren <slarengh@gmail.com> * fix llama_split_prefix --------- Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Xuan Son Nguyen <thichthat@gmail.com>
2024-03-22 18:00:01 +00:00
return size_prefix;
}
return 0;
}
Expose generation timings from server & update completions.js (#2116) * use javascript generators as much cleaner API Also add ways to access completion as promise and EventSource * export llama_timings as struct and expose them in server * update readme, update baked includes * llama : uniform variable names + struct init --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-05 20:51:13 +00:00
struct llama_timings llama_get_timings(struct llama_context * ctx) {
struct llama_timings result = {
/*.t_start_ms =*/ 1e-3 * ctx->t_start_us,
/*.t_end_ms =*/ 1.00 * ggml_time_ms(),
/*.t_load_ms =*/ 1e-3 * ctx->t_load_us,
/*.t_sample_ms =*/ 1e-3 * ctx->t_sample_us,
/*.t_p_eval_ms =*/ 1e-3 * ctx->t_p_eval_us,
/*.t_eval_ms =*/ 1e-3 * ctx->t_eval_us,
/*.n_sample =*/ std::max(1, ctx->n_sample),
/*.n_p_eval =*/ std::max(1, ctx->n_p_eval),
/*.n_eval =*/ std::max(1, ctx->n_eval),
};
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
Expose generation timings from server & update completions.js (#2116) * use javascript generators as much cleaner API Also add ways to access completion as promise and EventSource * export llama_timings as struct and expose them in server * update readme, update baked includes * llama : uniform variable names + struct init --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-05 20:51:13 +00:00
return result;
}
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
Expose generation timings from server & update completions.js (#2116) * use javascript generators as much cleaner API Also add ways to access completion as promise and EventSource * export llama_timings as struct and expose them in server * update readme, update baked includes * llama : uniform variable names + struct init --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-05 20:51:13 +00:00
void llama_print_timings(struct llama_context * ctx) {
const llama_timings timings = llama_get_timings(ctx);
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
add log_callback to llama_context_params for custom logging. (#2234) * add log_callback to llama_context_params for custom logging. * Fix macro expansion on gcc * Add struct llama_state for global variables and move log_callback there * Turn log level into enum and some minor changes. * Remove model_for_logging parameter (not needed anymore) * Convert remaining fprintf(stderr, ...) calls to use new macros. * Fix enum and initialize g_state * Fix log calls after merge * Fix missing static * Add back all the new lines in the logging strings * Add comment for llama_log_callback and replace remaining printf calls --------- Co-authored-by: grahameth <-> Co-authored-by: Helmut <helmut.buhler@inf.h-brs.de>
2023-08-09 20:46:40 +00:00
LLAMA_LOG_INFO("\n");
parallel : add option to load external prompt file (#3416) * Enable external file and add datestamp * Add name of external file at end * Upload ToK2024 * Delete ToK2024.txt * Experiments with jeopardy * Move ParallelQuestions to /proimpts and rename * Interim commit * Interim commit * Final revision * Remove trailing whitespace * remove cmake_all.sh * Remove cmake_all.sh * Changed .gitignore * Improved reporting and new question files. * Corrected typo * More LLM questions * Update LLM-questions.txt * Yet more LLM-questions * Remove jeopardy results file * Reinstate original jeopardy.sh * Update examples/parallel/parallel.cpp --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-06 13:16:38 +00:00
LLAMA_LOG_INFO("%s: load time = %10.2f ms\n", __func__, timings.t_load_ms);
LLAMA_LOG_INFO("%s: sample time = %10.2f ms / %5d runs (%8.2f ms per token, %8.2f tokens per second)\n",
Expose generation timings from server & update completions.js (#2116) * use javascript generators as much cleaner API Also add ways to access completion as promise and EventSource * export llama_timings as struct and expose them in server * update readme, update baked includes * llama : uniform variable names + struct init --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-05 20:51:13 +00:00
__func__, timings.t_sample_ms, timings.n_sample, timings.t_sample_ms / timings.n_sample, 1e3 / timings.t_sample_ms * timings.n_sample);
parallel : add option to load external prompt file (#3416) * Enable external file and add datestamp * Add name of external file at end * Upload ToK2024 * Delete ToK2024.txt * Experiments with jeopardy * Move ParallelQuestions to /proimpts and rename * Interim commit * Interim commit * Final revision * Remove trailing whitespace * remove cmake_all.sh * Remove cmake_all.sh * Changed .gitignore * Improved reporting and new question files. * Corrected typo * More LLM questions * Update LLM-questions.txt * Yet more LLM-questions * Remove jeopardy results file * Reinstate original jeopardy.sh * Update examples/parallel/parallel.cpp --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-06 13:16:38 +00:00
LLAMA_LOG_INFO("%s: prompt eval time = %10.2f ms / %5d tokens (%8.2f ms per token, %8.2f tokens per second)\n",
Expose generation timings from server & update completions.js (#2116) * use javascript generators as much cleaner API Also add ways to access completion as promise and EventSource * export llama_timings as struct and expose them in server * update readme, update baked includes * llama : uniform variable names + struct init --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-05 20:51:13 +00:00
__func__, timings.t_p_eval_ms, timings.n_p_eval, timings.t_p_eval_ms / timings.n_p_eval, 1e3 / timings.t_p_eval_ms * timings.n_p_eval);
parallel : add option to load external prompt file (#3416) * Enable external file and add datestamp * Add name of external file at end * Upload ToK2024 * Delete ToK2024.txt * Experiments with jeopardy * Move ParallelQuestions to /proimpts and rename * Interim commit * Interim commit * Final revision * Remove trailing whitespace * remove cmake_all.sh * Remove cmake_all.sh * Changed .gitignore * Improved reporting and new question files. * Corrected typo * More LLM questions * Update LLM-questions.txt * Yet more LLM-questions * Remove jeopardy results file * Reinstate original jeopardy.sh * Update examples/parallel/parallel.cpp --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-10-06 13:16:38 +00:00
LLAMA_LOG_INFO("%s: eval time = %10.2f ms / %5d runs (%8.2f ms per token, %8.2f tokens per second)\n",
Expose generation timings from server & update completions.js (#2116) * use javascript generators as much cleaner API Also add ways to access completion as promise and EventSource * export llama_timings as struct and expose them in server * update readme, update baked includes * llama : uniform variable names + struct init --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-07-05 20:51:13 +00:00
__func__, timings.t_eval_ms, timings.n_eval, timings.t_eval_ms / timings.n_eval, 1e3 / timings.t_eval_ms * timings.n_eval);
main : print total token count and tokens consumed so far (#4874) * Token count changes * Add show token count * Updating before PR * Two requested changes * Move param def posn
2024-01-11 16:14:52 +00:00
LLAMA_LOG_INFO("%s: total time = %10.2f ms / %5d tokens\n", __func__, (timings.t_end_ms - timings.t_start_ms), (timings.n_p_eval + timings.n_eval));
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
void llama_reset_timings(struct llama_context * ctx) {
ctx->t_start_us = ggml_time_us();
ctx->t_sample_us = ctx->n_sample = 0;
ctx->t_eval_us = ctx->n_eval = 0;
Add timings for the prompt evaluation (#478)
2023-03-25 14:34:23 +00:00
ctx->t_p_eval_us = ctx->n_p_eval = 0;
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
}
const char * llama_print_system_info(void) {
static std::string s;
s = "";
Speedup the AVX-512 implementation of ggml_vec_dot_q4_0() (#933)
2023-04-17 13:10:57 +00:00
s += "AVX = " + std::to_string(ggml_cpu_has_avx()) + " | ";
ggml : add ggml_cpu_has_avx_vnni() (#4589) * feat: add avx_vnni based on intel documents * ggml: add avx vnni based on intel document * llama: add avx vnni information display * docs: add more details about using oneMKL and oneAPI for intel processors * docs: add more details about using oneMKL and oneAPI for intel processors * docs: add more details about using oneMKL and oneAPI for intel processors * docs: add more details about using oneMKL and oneAPI for intel processors * docs: add more details about using oneMKL and oneAPI for intel processors * Update ggml.c Fix indentation upgate Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-12-30 08:07:48 +00:00
s += "AVX_VNNI = " + std::to_string(ggml_cpu_has_avx_vnni()) + " | ";
Speedup the AVX-512 implementation of ggml_vec_dot_q4_0() (#933)
2023-04-17 13:10:57 +00:00
s += "AVX2 = " + std::to_string(ggml_cpu_has_avx2()) + " | ";
s += "AVX512 = " + std::to_string(ggml_cpu_has_avx512()) + " | ";
s += "AVX512_VBMI = " + std::to_string(ggml_cpu_has_avx512_vbmi()) + " | ";
s += "AVX512_VNNI = " + std::to_string(ggml_cpu_has_avx512_vnni()) + " | ";
s += "FMA = " + std::to_string(ggml_cpu_has_fma()) + " | ";
s += "NEON = " + std::to_string(ggml_cpu_has_neon()) + " | ";
s += "ARM_FMA = " + std::to_string(ggml_cpu_has_arm_fma()) + " | ";
s += "F16C = " + std::to_string(ggml_cpu_has_f16c()) + " | ";
s += "FP16_VA = " + std::to_string(ggml_cpu_has_fp16_va()) + " | ";
s += "WASM_SIMD = " + std::to_string(ggml_cpu_has_wasm_simd()) + " | ";
s += "BLAS = " + std::to_string(ggml_cpu_has_blas()) + " | ";
s += "SSE3 = " + std::to_string(ggml_cpu_has_sse3()) + " | ";
ggml : detect SSSE3 (#2825) * ggml : add ggml_cpu_has_ssse3 * llama : show SSSE3 in system info
2023-08-27 08:10:25 +00:00
s += "SSSE3 = " + std::to_string(ggml_cpu_has_ssse3()) + " | ";
Speedup the AVX-512 implementation of ggml_vec_dot_q4_0() (#933)
2023-04-17 13:10:57 +00:00
s += "VSX = " + std::to_string(ggml_cpu_has_vsx()) + " | ";
ggml : add mmla kernels for quantized GEMM (#4966) * ggml: aarch64: implement smmla kernel for q8_0_q8_0 quantized gemm armv8.2-a and above supports MMLA instructions that have higher throughput than DOT. this commit adds mmla kernel for q8_0_q8_0 gemm. The feature is enabled if the platform supports "__ARM_FEATURE_MATMUL_INT8" On AWS Graviton3 processors this kernel resulted up to 1.5x improvement for prompt evaluation throughput compared to the default sdot kernel. * ggml: aarch64: implement smmla kernel for q4_0_q8_0 quantized gemm armv8.2-a and above supports MMLA instructions that have higher throughput than DOT. this commit adds mmla kernel for q4_0_q8_0 gemm. The feature is enabled if the platform supports "__ARM_FEATURE_MATMUL_INT8" On AWS Graviton3 processors this kernel resulted up to 1.5x improvement for prompt evaluation throughput compared to the default sdot kernel. * ggml: aarch64: implement smmla kernel for q4_1_q8_1 quantized gemm armv8.2-a and above supports MMLA instructions that have higher throughput than DOT. this commit adds mmla kernel for q4_1_q8_1 gemm. The feature is enabled if the platform supports "__ARM_FEATURE_MATMUL_INT8" On AWS Graviton3 processors this kernel resulted up to 1.5x improvement for prompt evaluation throughput compared to the default sdot kernel. * ggml: update unit tests for the new vec_dot interface * llama.cpp: add MATMUL_INT8 capability to system_info
2024-02-11 13:22:33 +00:00
s += "MATMUL_INT8 = " + std::to_string(ggml_cpu_has_matmul_int8()) + " | ";
cmake : restore LLAMA_LLAMAFILE_DEFAULT
2024-04-25 18:31:17 +00:00
#ifdef GGML_USE_LLAMAFILE
llama : fix typo LAMMAFILE -> LLAMAFILE (#6974)
2024-04-29 12:36:22 +00:00
s += "LLAMAFILE = 1 | ";
cmake : restore LLAMA_LLAMAFILE_DEFAULT
2024-04-25 18:31:17 +00:00
#else
llama : fix typo LAMMAFILE -> LLAMAFILE (#6974)
2024-04-29 12:36:22 +00:00
s += "LLAMAFILE = 0 | ";
cmake : restore LLAMA_LLAMAFILE_DEFAULT
2024-04-25 18:31:17 +00:00
#endif
Introduce C-style API (#370) * Major refactoring - introduce C-style API * Clean up * Add <cassert> * Add <iterator> * Add <algorithm> .... * Fix timing reporting and accumulation * Measure eval time only for single-token calls * Change llama_tokenize return meaning
2023-03-22 05:32:36 +00:00
return s.c_str();
}
Add quantize-stats command for testing quantization (#728) Command that calculates some statistics over the errors introduced by quantization, like mean square error, max error and some percentile errors for layer weights. Should be useful for testing quantization improvements. Exposes some internal state from ggml and llama for testing
2023-04-07 22:09:18 +00:00
YAML result logging + preset script (#2657)
2023-08-28 15:59:39 +00:00
void llama_dump_timing_info_yaml(FILE * stream, const llama_context * ctx) {
fprintf(stream, "\n");
fprintf(stream, "###########\n");
fprintf(stream, "# Timings #\n");
fprintf(stream, "###########\n");
fprintf(stream, "\n");
fprintf(stream, "mst_eval: %.2f # ms / token during generation\n",
1.0e-3 * ctx->t_eval_us / ctx->n_eval);
fprintf(stream, "mst_p_eval: %.2f # ms / token during prompt processing\n",
1.0e-3 * ctx->t_p_eval_us / ctx->n_p_eval);
fprintf(stream, "mst_sample: %.2f # ms / token during sampling\n",
1.0e-3 * ctx->t_sample_us / ctx->n_sample);
fprintf(stream, "n_eval: %d # number of tokens generated (excluding the first one)\n", ctx->n_eval);
fprintf(stream, "n_p_eval: %d # number of tokens processed in batches at the beginning\n", ctx->n_p_eval);
fprintf(stream, "n_sample: %d # number of sampled tokens\n", ctx->n_sample);
train : mem usage and other improvements (#2439) * fix track_max_mem in forward_batch_wo_cache_flash_attn_train * remove unnecessary Adam(W) optimizer tensors. reduces optimizer memory overhead from 7*modelsize to 2*modelsize. additionally allows to optimize models with more than 2^31 parameters by replacing int with int64_t. bumps training checkpoint file version, but old checkpoints can still be read. new version with less tensors is saved. * add gradient clipping to AdamW * Fix reset of unused g->nodes and g->grads to NULL * implement gradient checkpointing for training reduces memory overhead from O(n_layer) to O(sqrt(n_layer)) as explained in readme of https://github.com/cybertronai/gradient-checkpointing * remove unused compute buffer 3 * add and use function ggml_build_backward_expand to avoid stack overflows with large maximum number of nodes GGML_API void ggml_build_backward_expand(struct ggml_context * ctx, struct ggml_cgraph * gf, struct ggml_cgraph * gb, bool keep); * change AdamW decay parameter to work like the torch AdamW decay parameter It is now relative to Adam learning rate `alpha*sched`. Before that it was relative to `sched` only. `alpha` being the maximum learning rate and `sched` being a scaling parameter in [0..1] * change default AdamW weight decay parameter used in training to 0.1 as used in nanoGPT * change default AdamW weight decay parameter defined in ggml to 0.0, making Adam default instead of AdamW btw: the default weight decay parameter for torch.optim.AdamW is 0.01 * bug fixes for cross entropy loss ggml_cross_entropy_loss: sums where not correctly added in workload of each thread ggml_cross_entropy_loss_back: simplify backward process, reducing numerical issues guard usage of exp f16 lookup in cross entropy by #define GGML_CROSS_ENTROPY_EXP_FP16 cross entropy loss is only used once during training, but it is quite sensitive to numerical errors introduced by exp-f16-lookup. so exp-f16-lookup for cross entropy loss is disabled by default, trading better gradients for very slightly worse runtime performance. * fix test-grad0 for cross_entropy_loss the second argument to cross_entropy_loss must sum up to 1 for each row * fix test-grad0 for soft_max dont use only sum as aggregation, because sum of softmax is always 1 -> finite differences should not work instead use sum(log(soft_max()*(1-eps)+eps)); use eps to avoid log(0) * improve finite differences of test-grad0 by using double instead of float * change cross_entropy_loss to output average over all rows this helps keeping the loss and gradients in a sane range * improve gradient checkpointing sqrt(n_layers) is only the best checkpoint step when mem size of checkpoints and mem size of layers are equal. since layers require more memory than the single-tensor-checkpoint we use, the optimal values are compute different: ``` given: n, u, v objective: minimize(a*u+b*v) where a*b=n, a>0, b>0 b=n/a minimize(a*u+v*n/a) diff(a*u+v*n/a, a) = u - (v*n/a)/a diff(a*u+v*n/a, a) == 0 u - (v*n/a)/a == 0 u == v*n/(a*a) u*a*a = v*n a*a = v*n/u a = sqrt(n*v/u) ``` this change results in more checkpoints, requiring less layers to store between checkpoints, overall improving memory usage. * disable gradient checkpointing debug output * llama : fix rope usage in train-text-from-scratch after ChatGLM change * add more training parameters: --enable-restart N Only for Adam optimizer. Enable restarts of cos-decay --disable-restart N Only for Adam optimizer. Disable restarts of cos-decay --opt-past N Number of optimization iterations to track for delta convergence test. Disabled when zero. --opt-delta N Maximum delta for delta convergence test. Disabled when <= zero. --opt-max-no-improvement N Maximum number of optimization iterations with no improvement. Disabled when <= zero. --adam-epsf N AdamW epsilon for convergence test. Disabled when <= zero. --adam-min-alpha N Adam minimum learning rate alpha, usually 0.1 * alpha * replace memcpy with reshape operation so that the graph is not cut at the input this makes it possible to store other values into the input tensor and then simply recompute the graph without rebuilding it * remove unused function argument from get_example_targets_batch * measure and print total training time * add optimization callback to ggml_opt_resume_g this callback is called before each iteration with custom data and pointer to learning schedule parameter (only used in Adam(W)). can be used for dynamic learning schedule and setting input data for batches before each iteration * use optimization callback in training allows dynamic learning schedule and different batch data for each iteration without relying on low n_iter and high n_examples parameters reduces runtime by avoiding restart of optimization function and improves training convergence by providing a different batch for each iteration * add minimum number of tensor dimensions to apply weight decay (default 2) this allows to not apply weight decay to bias parameters * rename training parameter cos-decay-alpha to cos-decay-min and clarify that adam-min-alpha also applies to warmup * fix increase of model.train_samples and model.train_tokens now that each optimizer iteration gets its own batch we need to multiply by number of opt iterations * change sampling parameters for prediction after training to defaults of common.h and clarify what is context for prediction and what are generated tokens * tighten abs error bounds for cross_entropy_loss in test-grad0 * add conditional compilation of using F16 exp in flash attention uncomment `// #define GGML_FLASH_ATTN_EXP_FP16` to enable usage of f16 exp in flash attention * tighten abs error bounds for flash_attn in test-grad0 * tighten abs error bounds for sqrt in test-grad0 * remove out-commented vectorized code of opt_adam the vectorized code might be bit faster for low number of parameters, but it had a big memory usage overhead * ggml : update ggml_rms_norm_back with configurable eps * llama training : fix ggml_rms_norm_back calls to pass configurable eps * remove trailing whitespace * add train function using automatic gradient checkpointing backward pass and allocator * in train function replace add_inplace by regular add because using add_inplace seems to result in different gradients * don't use allocate hash_map on context because the context has no_alloc=True when using memory allocator resulting in NULL data pointers * correctly clone reshape and permute operations by also cloning tensor->nb values * fix variable name and add missing type cast * terminate recursive tensor cloning when reaching tensor without src tensors * correctly clone view tensors by setting data pointers without this the checkpointing would only work when being used together with memory allocator * fix variable names * swap arguments to commutative ops to be the same as in `forward_batch_wo_cache_flash_attn` * add input tensors as checkpoints so that recursive tensor cloning of gradient checkpointing terminates on input tensors * fix variable name and add missing boolean negation * make sure some tensors are not reallocated by inserting new temporary nodes depending on them: output and parameter gradient tensors need to be available at the end of the graph execution parameter gradient tensors also need to be available before the graph execution because they are set to zero before each optimizer iteration checkpoint tensors are allocated all together to reduce memory allocator fragmentation afterwards, in addition to the temporary nodes, we also need to reset the temporary leafs * fix ASSERT to work with zero layers * add training options whether to use allocator and/or unified training function * integrate unified training function which may use memory allocator the unified training function also supports arguments whether to use flash attention and/or gradient checkpointing * format name of cloned tensors with " (clone)" suffix * set names for tensors in unified train function for easier debugging * allocate graph on context using ggml_new_graph * remove handwritten training functions * remove unused training parameters "use_scratch" and "use_unified" * remove trailing whitespace * remove unused train params: mem_compute1_gb & mem_compute2_gb mem_compute_gb is used for compute when automatic memory allocator is not enabled, otherwise it can be very small to only hold the tensor definitions mem_compute0_gb is used for automatic memory allocator (as long as measurement of max required size is not implemented) * remove unused forward_batch function * add debug asserts in ggml_allocr_alloc to some common pitfalls when using this function directly * only use ggml_allocr_alloc when tensor has NULL data and is no view * fix test when to create temporary backward graph temporary backward graph is only necessary when using checkpointing * fix memory "leak" in optimizers each iteration a new cplan with new memory for work data was allocated. now cplan creation only happens at the start of optimization, with each iteration reusing the cplan and its work data. * reverse order of for loop in ggml_build_backward_expand to save memory when using gradient checkpointing and allocator with this loop order gradient checkpointing with allocator on 16 layer model saves 13% memory; 2 layer memory it saves 2% memory. the computation results are the same * add missing lctx argument to get_example_targets_batch * implement llama model file saving using gguf checkpoint loading and saving disabled, to be replaced by loading and saving via gguf * implement loading/saving of checkpointing files using GGUF * bug fixes * add checkpoint file version for future compatibility * update readme with gguf filenames * save & load opt->just_initialized value * add first draft for checkpoint conversion script * add gguf arch and ftype * save opt parameter counter as uint64 * add gguf key and tensor names for optimizer and training * add layer_norm_rms_eps to checkpoint convert script * use same GGUF_GET_KEY macro as in llama.cpp * use norm_rms_eps, and rope parameters and command line options to set them * fix memory corruption bug in gguf ctx->kv and ctx->infos was reallocated using not-aligned realloc, but freed with aligned free. to fix this a GGML_ALIGNED_REALLOC was added, but there is no posix_memalign_realloc function. so on non-windows and non-mingw32 platforms we fall back to aligned malloc, followed by copying and freeing the old data. * add gguf example cmake file * bug fixes in tokenize_file * bug fixes in load_llama_model_gguf * bug fix: init model when no checkpoint was loaded * bug fix in read_tensor_by_name * bug fix in load_opt_context_gguf * avoid printing lots of spaced on the unusual case that loss gets nan * set name of tensors with empty name from what was read from gguf * remove trailing whitespace * print data checksums before saving and after loading to verify correctness * bug fixes for convert-train-checkpoint-to-gguf * temporarily add code to write old checkpoint files used to verify that old checkpoint files are correctly converted to gguf * bug fixes for convert-train-checkpoint-to-gguf.py loading checkpoints with opt_version=0 * remove code used to verify correctness of checkpoint file conversion * remove trailing whitespace * remove prediction related code use main for prediction, it is better optimized * update train-text-from-scratch README.md * fix non-windows GGML_ALIGNED_REALLOC * add missing blank line at end of file * remove GGML_ALIGNED_REALLOC and use normal malloc/realloc/free for gguf ctx->kv & ctx->infos * train : fix compile warnings --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-08-28 19:51:47 +00:00
fprintf(stream, "t_eval_us: %" PRId64 " # total microseconds spent generating tokens\n", ctx->t_eval_us);
fprintf(stream, "t_load_us: %" PRId64 " # total microseconds spent loading the model\n", ctx->t_load_us);
fprintf(stream, "t_p_eval_us: %" PRId64 " # total microseconds spent prompt processing\n", ctx->t_p_eval_us);
fprintf(stream, "t_sample_us: %" PRId64 " # total microseconds spent sampling\n", ctx->t_sample_us);
YAML result logging + preset script (#2657)
2023-08-28 15:59:39 +00:00
fprintf(stream, "ts_eval: %.2f # tokens / second during generation\n",
1.0e6 * ctx->n_eval / ctx->t_eval_us);
fprintf(stream, "ts_p_eval: %.2f # tokens / second during prompt processing\n",
1.0e6 * ctx->n_p_eval / ctx->t_p_eval_us);
fprintf(stream, "ts_sample: %.2f # tokens / second during sampling\n",
1.0e6 * ctx->n_sample / ctx->t_sample_us);
}
Add quantize-stats command for testing quantization (#728) Command that calculates some statistics over the errors introduced by quantization, like mean square error, max error and some percentile errors for layer weights. Should be useful for testing quantization improvements. Exposes some internal state from ggml and llama for testing
2023-04-07 22:09:18 +00:00
// For internal test use
check C++ code with -Wmissing-declarations (#3184)
2023-09-15 19:38:27 +00:00
const std::vector<std::pair<std::string, struct ggml_tensor *>> & llama_internal_get_tensor_map(
struct llama_context * ctx
) {
Rewrite loading code to try to satisfy everyone: - Support all three formats (ggml, ggmf, ggjt). (However, I didn't include the hack needed to support GPT4All files without conversion. Those can still be used after converting them with convert.py from my other PR.) - Support both mmap and read (mmap is used by default, but can be disabled with `--no-mmap`, and is automatically disabled for pre-ggjt files or on platforms where mmap is not supported). - Support multi-file models like before, but automatically determine the number of parts rather than requiring `--n_parts`. - Improve validation and error checking. - Stop using the per-file type field (f16) entirely in favor of just relying on the per-tensor type/size fields. This has no immediate benefit, but makes it easier to experiment with different formats, and should make it easier to support the new GPTQ-for-LLaMa models in the future (I have some work in progress on that front). - Support VirtualLock on Windows (using the same `--mlock` option as on Unix). - Indicate loading progress when using mmap + mlock. (Which led me to the interesting observation that on my Linux machine, with a warm file cache, mlock actually takes some time, whereas mmap without mlock starts almost instantly...) - To help implement this, move mlock support from ggml to the loading code. - madvise/PrefetchVirtualMemory support (based on #740) - Switch from ifstream to the `fopen` family of functions to avoid unnecessary copying and, when mmap is enabled, allow reusing the same file descriptor for both metadata reads and mmap (whereas the existing implementation opens the file a second time to mmap). - Quantization now produces a single-file output even with multi-file inputs (not really a feature as much as 'it was easier this way'). Implementation notes: I tried to factor the code into more discrete pieces than before. Regarding code style: I tried to follow the code style, but I'm naughty and used a few advanced C++ features repeatedly: - Destructors to make it easier to ensure everything gets cleaned up. - Exceptions. I don't even usually use exceptions when writing C++, and I can remove them if desired... but here they make the loading code much more succinct while still properly handling a variety of errors, ranging from API calls failing to integer overflow and allocation failure. The exceptions are converted to error codes at the API boundary.) Co-authored-by: Pavol Rusnak <pavol@rusnak.io> (for the bit I copied from #740)
2023-04-08 19:24:37 +00:00
return ctx->model.tensors_by_name;
Add quantize-stats command for testing quantization (#728) Command that calculates some statistics over the errors introduced by quantization, like mean square error, max error and some percentile errors for layer weights. Should be useful for testing quantization improvements. Exposes some internal state from ggml and llama for testing
2023-04-07 22:09:18 +00:00
}
add log_callback to llama_context_params for custom logging. (#2234) * add log_callback to llama_context_params for custom logging. * Fix macro expansion on gcc * Add struct llama_state for global variables and move log_callback there * Turn log level into enum and some minor changes. * Remove model_for_logging parameter (not needed anymore) * Convert remaining fprintf(stderr, ...) calls to use new macros. * Fix enum and initialize g_state * Fix log calls after merge * Fix missing static * Add back all the new lines in the logging strings * Add comment for llama_log_callback and replace remaining printf calls --------- Co-authored-by: grahameth <-> Co-authored-by: Helmut <helmut.buhler@inf.h-brs.de>
2023-08-09 20:46:40 +00:00
metal : reusing llama.cpp logging (#3152) * metal : reusing llama.cpp logging * cmake : build fix * metal : logging callback * metal : logging va_args memory fix * metal : minor cleanup * metal : setting function like logging macro to capital letters * llama.cpp : trailing whitespace fix * ggml : log level enum used by llama * Makefile : cleanup ggml-metal recipe * ggml : ggml_log_callback typedef * ggml : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-27 15:48:33 +00:00
void llama_log_set(ggml_log_callback log_callback, void * user_data) {
add log_callback to llama_context_params for custom logging. (#2234) * add log_callback to llama_context_params for custom logging. * Fix macro expansion on gcc * Add struct llama_state for global variables and move log_callback there * Turn log level into enum and some minor changes. * Remove model_for_logging parameter (not needed anymore) * Convert remaining fprintf(stderr, ...) calls to use new macros. * Fix enum and initialize g_state * Fix log calls after merge * Fix missing static * Add back all the new lines in the logging strings * Add comment for llama_log_callback and replace remaining printf calls --------- Co-authored-by: grahameth <-> Co-authored-by: Helmut <helmut.buhler@inf.h-brs.de>
2023-08-09 20:46:40 +00:00
g_state.log_callback = log_callback ? log_callback : llama_log_callback_default;
g_state.log_callback_user_data = user_data;
llama : set metal log callback correctly (#4204)
2023-11-24 17:10:01 +00:00
#ifdef GGML_USE_METAL
metal : remove old API (#4919) ggml-ci
2024-01-13 18:45:45 +00:00
ggml_backend_metal_log_set_callback(g_state.log_callback, g_state.log_callback_user_data);
llama : set metal log callback correctly (#4204)
2023-11-24 17:10:01 +00:00
#endif
add log_callback to llama_context_params for custom logging. (#2234) * add log_callback to llama_context_params for custom logging. * Fix macro expansion on gcc * Add struct llama_state for global variables and move log_callback there * Turn log level into enum and some minor changes. * Remove model_for_logging parameter (not needed anymore) * Convert remaining fprintf(stderr, ...) calls to use new macros. * Fix enum and initialize g_state * Fix log calls after merge * Fix missing static * Add back all the new lines in the logging strings * Add comment for llama_log_callback and replace remaining printf calls --------- Co-authored-by: grahameth <-> Co-authored-by: Helmut <helmut.buhler@inf.h-brs.de>
2023-08-09 20:46:40 +00:00
}
metal : reusing llama.cpp logging (#3152) * metal : reusing llama.cpp logging * cmake : build fix * metal : logging callback * metal : logging va_args memory fix * metal : minor cleanup * metal : setting function like logging macro to capital letters * llama.cpp : trailing whitespace fix * ggml : log level enum used by llama * Makefile : cleanup ggml-metal recipe * ggml : ggml_log_callback typedef * ggml : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-27 15:48:33 +00:00
static void llama_log_internal_v(ggml_log_level level, const char * format, va_list args) {
add log_callback to llama_context_params for custom logging. (#2234) * add log_callback to llama_context_params for custom logging. * Fix macro expansion on gcc * Add struct llama_state for global variables and move log_callback there * Turn log level into enum and some minor changes. * Remove model_for_logging parameter (not needed anymore) * Convert remaining fprintf(stderr, ...) calls to use new macros. * Fix enum and initialize g_state * Fix log calls after merge * Fix missing static * Add back all the new lines in the logging strings * Add comment for llama_log_callback and replace remaining printf calls --------- Co-authored-by: grahameth <-> Co-authored-by: Helmut <helmut.buhler@inf.h-brs.de>
2023-08-09 20:46:40 +00:00
va_list args_copy;
va_copy(args_copy, args);
char buffer[128];
int len = vsnprintf(buffer, 128, format, args);
if (len < 128) {
g_state.log_callback(level, buffer, g_state.log_callback_user_data);
} else {
char* buffer2 = new char[len+1];
vsnprintf(buffer2, len+1, format, args_copy);
buffer2[len] = 0;
g_state.log_callback(level, buffer2, g_state.log_callback_user_data);
delete[] buffer2;
}
va_end(args_copy);
}
metal : reusing llama.cpp logging (#3152) * metal : reusing llama.cpp logging * cmake : build fix * metal : logging callback * metal : logging va_args memory fix * metal : minor cleanup * metal : setting function like logging macro to capital letters * llama.cpp : trailing whitespace fix * ggml : log level enum used by llama * Makefile : cleanup ggml-metal recipe * ggml : ggml_log_callback typedef * ggml : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-27 15:48:33 +00:00
static void llama_log_internal(ggml_log_level level, const char * format, ...) {
add log_callback to llama_context_params for custom logging. (#2234) * add log_callback to llama_context_params for custom logging. * Fix macro expansion on gcc * Add struct llama_state for global variables and move log_callback there * Turn log level into enum and some minor changes. * Remove model_for_logging parameter (not needed anymore) * Convert remaining fprintf(stderr, ...) calls to use new macros. * Fix enum and initialize g_state * Fix log calls after merge * Fix missing static * Add back all the new lines in the logging strings * Add comment for llama_log_callback and replace remaining printf calls --------- Co-authored-by: grahameth <-> Co-authored-by: Helmut <helmut.buhler@inf.h-brs.de>
2023-08-09 20:46:40 +00:00
va_list args;
va_start(args, format);
llama_log_internal_v(level, format, args);
va_end(args);
}
metal : reusing llama.cpp logging (#3152) * metal : reusing llama.cpp logging * cmake : build fix * metal : logging callback * metal : logging va_args memory fix * metal : minor cleanup * metal : setting function like logging macro to capital letters * llama.cpp : trailing whitespace fix * ggml : log level enum used by llama * Makefile : cleanup ggml-metal recipe * ggml : ggml_log_callback typedef * ggml : minor --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2023-09-27 15:48:33 +00:00
static void llama_log_callback_default(ggml_log_level level, const char * text, void * user_data) {
add log_callback to llama_context_params for custom logging. (#2234) * add log_callback to llama_context_params for custom logging. * Fix macro expansion on gcc * Add struct llama_state for global variables and move log_callback there * Turn log level into enum and some minor changes. * Remove model_for_logging parameter (not needed anymore) * Convert remaining fprintf(stderr, ...) calls to use new macros. * Fix enum and initialize g_state * Fix log calls after merge * Fix missing static * Add back all the new lines in the logging strings * Add comment for llama_log_callback and replace remaining printf calls --------- Co-authored-by: grahameth <-> Co-authored-by: Helmut <helmut.buhler@inf.h-brs.de>
2023-08-09 20:46:40 +00:00
(void) level;
(void) user_data;
fputs(text, stderr);
fflush(stderr);
}
Reference in New Issue Copy Permalink